From 4e690c0e336a4e360ba96f4edf9e8da2c6d2b343 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 03:39:46 +0200 Subject: [PATCH 01/21] WIP: FastHashMap as a compiler-side proof MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Exercises type class bounds through the container they were added for. Six of the seven cases pass on both backends with no compiler change: int keys, tuple keys on Jass, a user class key, two specialisations coexisting, and two instances of one specialisation. tupleKeyLua fails and is a real, pre-existing backend bug. Method names become Lua table keys, but luaMethod.initFor passes the name through raw while every sibling (luaVar, luaFunc, luaClassVar) sanitises via uniqueName. Names are valid identifiers in ordinary code, so nothing hit it until a class method was specialised for Lua with more than one type argument: specializeMethod builds name + "_specialized_" + generics.makeName(), and makeName joins arguments with ", ". Two simple arguments give "get_specialized_integer, integer", which emits "Class.get_specialized_integer, integer = impl" -- valid Lua that assigns to two targets and quietly writes a junk global. A tuple argument gives "⦅integer, integer⦆" and fails the syntax check outright. --- .../wurstscript/tests/FastHashMapTests.java | 207 ++++++++++++++++++ 1 file changed, 207 insertions(+) create mode 100644 de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java new file mode 100644 index 000000000..50becb1f3 --- /dev/null +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java @@ -0,0 +1,207 @@ +package tests.wurstscript.tests; + +import org.testng.annotations.Test; + +/** + * Exercises type class bounds through the container they were added for: a hash map whose key type + * is bounded by {@code Hashable}, so it can hash and compare keys without erasing them to int. + *

+ * This is deliberately a whole working container rather than a minimal repro. It combines things + * nothing else tests together: two type parameters where only the first is bounded, static arrays of + * a bounded type parameter, and a bound used from a private method reached through a public one. + */ +public class FastHashMapTests extends WurstScriptTest { + + /** + * Storage follows ArrayList: one array per specialisation, carved into a section per instance. + * Collision handling is linear probing, so a lookup calls both requirements of the bound. + */ + private static String[] fastHashMap(String... extra) { + String[] head = { + "package test", + "native testSuccess()", + "constant int CAPACITY = 8", + "interface Hashable", + " function hash(T x) returns int", + " function equals(T a, T b) returns boolean", + "class FastHashMap", + " private static K array keys", + " private static V array values", + " private static boolean array used", + " private static int nextFree = 0", + " private int base", + " private int count = 0", + " construct()", + " base = nextFree", + " nextFree += CAPACITY", + " private function slotFor(K key) returns int", + " var i = K.hash(key) mod CAPACITY", + " if i < 0", + " i += CAPACITY", + " while used[base + i] and not K.equals(keys[base + i], key)", + " i = (i + 1) mod CAPACITY", + " return base + i", + " function put(K key, V value)", + " let s = slotFor(key)", + " if not used[s]", + " used[s] = true", + " keys[s] = key", + " count++", + " values[s] = value", + " function get(K key) returns V", + " return values[slotFor(key)]", + " function has(K key) returns boolean", + " return used[slotFor(key)]", + " function size() returns int", + " return count", + }; + String[] all = new String[head.length + extra.length]; + System.arraycopy(head, 0, all, 0, head.length); + System.arraycopy(extra, 0, all, head.length, extra.length); + return all; + } + + private static final String[] INT_INSTANCE = { + "implements Hashable", + " function hash(int x) returns int", + " return x", + " function equals(int a, int b) returns boolean", + " return a == b", + }; + + private static String[] program(String[]... parts) { + int size = 0; + for (String[] part : parts) { + size += part.length; + } + String[] all = new String[size]; + int at = 0; + for (String[] part : parts) { + System.arraycopy(part, 0, all, at, part.length); + at += part.length; + } + return all; + } + + /** Keys 1 and 9 land in the same slot at capacity 8, so the probe path is taken. */ + private static final String[] USE_WITH_COLLISION = { + "init", + " let m = new FastHashMap()", + " m.put(1, 10)", + " m.put(9, 90)", + " m.put(2, 20)", + " if m.get(1) == 10 and m.get(9) == 90 and m.get(2) == 20", + " if m.size() == 3 and not m.has(3)", + " testSuccess()", + }; + + @Test + public void fastHashMapRuntime() { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, USE_WITH_COLLISION)); + } + + @Test + public void fastHashMapRuntimeLua() { + test().testLua(true).executeProg().lines(program(fastHashMap(), INT_INSTANCE, USE_WITH_COLLISION)); + } + + /** + * A tuple key, which is the case old generics cannot serve at all: a tuple has no int + * representation to cast to, so the only way to key a map by one is to say how it hashes. + */ + private static final String[] TUPLE_INSTANCE = { + "tuple pos(int x, int y)", + "implements Hashable", + " function hash(pos p) returns int", + " return p.x * 31 + p.y", + " function equals(pos a, pos b) returns boolean", + " return a.x == b.x and a.y == b.y", + }; + + @Test + public void tupleKey() { + testAssertOkLines(true, program(fastHashMap(), TUPLE_INSTANCE, new String[]{ + "init", + " let m = new FastHashMap()", + " m.put(pos(1, 2), 12)", + " m.put(pos(2, 1), 21)", + " if m.get(pos(1, 2)) == 12 and m.get(pos(2, 1)) == 21", + " if m.has(pos(1, 2)) and not m.has(pos(9, 9))", + " testSuccess()" + })); + } + + @Test + public void tupleKeyLua() { + test().testLua(true).executeProg().lines(program(fastHashMap(), TUPLE_INSTANCE, new String[]{ + "init", + " let m = new FastHashMap()", + " m.put(pos(1, 2), 12)", + " m.put(pos(2, 1), 21)", + " if m.get(pos(1, 2)) == 12 and m.get(pos(2, 1)) == 21", + " testSuccess()" + })); + } + + /** A key type of the user's own, with the instance declared in the same package as the type. */ + @Test + public void classKeyWithUserInstance() { + testAssertOkLines(true, program(fastHashMap(), new String[]{ + "class Item", + " int id", + " construct(int id)", + " this.id = id", + "implements Hashable", + " function hash(Item i) returns int", + " return i.id", + " function equals(Item a, Item b) returns boolean", + " return a.id == b.id", + "init", + " let m = new FastHashMap()", + " let a = new Item(1)", + " let b = new Item(2)", + " m.put(a, \"a\")", + " m.put(b, \"b\")", + " if m.get(a) == \"a\" and m.get(b) == \"b\" and m.size() == 2", + " testSuccess()" + })); + } + + /** + * Two specialisations live at once. Storage is static per specialisation, so this is what would + * break if the arrays of one instantiation were shared with another. + */ + @Test + public void twoSpecialisationsCoexist() { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, new String[]{ + // Every string hashes alike, so this also exercises the probe path on every lookup. + "implements Hashable", + " function hash(string s) returns int", + " return 0", + " function equals(string a, string b) returns boolean", + " return a == b", + "init", + " let ints = new FastHashMap()", + " let strs = new FastHashMap()", + " ints.put(1, 100)", + " strs.put(\"a\", 200)", + " if ints.get(1) == 100 and strs.get(\"a\") == 200", + " if ints.size() == 1 and strs.size() == 1", + " testSuccess()" + })); + } + + /** Two maps of the same instantiation must not share storage either. */ + @Test + public void twoInstancesOfOneSpecialisation() { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, new String[]{ + "init", + " let a = new FastHashMap()", + " let b = new FastHashMap()", + " a.put(1, 10)", + " b.put(1, 20)", + " if a.get(1) == 10 and b.get(1) == 20 and a.size() == 1 and b.size() == 1", + " testSuccess()" + })); + } +} From 3ccbba245f652eb548ebfc047b6d0759ff79f7f3 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 03:44:47 +0200 Subject: [PATCH 02/21] Add a working backlog for the type class bound follow-ups --- BACKLOG.md | 70 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 70 insertions(+) create mode 100644 BACKLOG.md diff --git a/BACKLOG.md b/BACKLOG.md new file mode 100644 index 000000000..753288587 --- /dev/null +++ b/BACKLOG.md @@ -0,0 +1,70 @@ +# Language feature backlog + +Working notes for ongoing work on type class bounds for `T:` generics +(shipped in #1226, #1228, #1229). Ordered: take the top unblocked item. + +Keep this file current. It is the only memory that survives between sessions. +When an item is finished, move it to Done with one line on what actually +happened. When something is learned that would have saved time, write it under +Notes rather than leaving it in a commit message. + +## Todo + +1. **Lua method names reach the backend unsanitised.** `LuaTranslator.luaMethod.initFor` + passes `a.getName()` raw; `luaVar`, `luaFunc` and `luaClassVar` all sanitise via + `uniqueName`. Method names become Lua table keys, so they must be valid identifiers. + `EliminateGenerics.specializeMethod` builds `name + "_specialized_" + generics.makeName()` + on the Lua path, and `makeName` joins type arguments with `", "` — so a class method + specialised with two type arguments emits `Class.get_specialized_integer, integer = impl`, + which is valid Lua assigning to two targets, and a tuple argument emits `⦅⦆` and fails + the syntax check. Repro: `FastHashMapTests.tupleKeyLua`. Commas are already visible in + `test-output/lua/FastHashMapTests_fastHashMapRuntimeLua.lua`. + Overriding methods must keep landing in the same slot, so normalise per distinct original + name, not per method node. + +2. **`slotFor` looks bound to `get`'s implementation** in the same emitted Lua. May be a real + mis-binding in `specializeMethod`/`adaptSubmethods`, may be an artefact of item 1 mangling + the output. Diagnose only after item 1, from freshly emitted Lua. + +3. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. + Add `remove` with tombstones, and an assertion that the emitted code stays cheap: no + dispatch node, no instance dictionary, and no WC3 hashtable natives — array access only, + which is the whole point versus `HashMap extends Table`. + +4. **Lua dispatch inside a closure.** Works on Jass since #1229. On Lua the specialised class + is built correctly but nothing calls it, because the closure is reached through its + interface and `specializeMethod` renames the method out of its dispatch slot. + `TypeClassTests.dispatchInsideClosureIsRejectedForLua` pins the current diagnostic and + should become a success test. Related to item 1; AGENTS.md flags this machinery. + +5. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass. + +6. **Module bounds.** `module M` is rejected with a clear message today. Needs + receiver rewriting during expansion, or type parameters on `ModuleInstanciation`. + +7. **One `ImTypeVar` per type parameter.** Name-tolerant lookups remain in + `EliminateGenerics.indexOfTypeVar`, `inheritTypeClassBinding` and + `ProgramState.getCurrentTypeArgument`, compensating for several nodes standing for one + source parameter. Making the node canonical lets all three compare by identity and removes + a class of silent wrong dispatch. Mechanical, well covered by the suite. + +8. **Jass temp counter is not reset between compilations.** Two runs of the same commit emit + different `.j` (`temp151` vs `temp8`) because the counter is JVM-wide and depends on how + many tests ran before. Not wrong for compiling one map, but it means `.j` cannot be diffed + across runs to validate a change — only `.lua` can. Fixing it would make Jass diffable. + +## Done + +- Substitution now carries the type class binding with the type (#1229). Also fixed the + type-variable reference on `ImTypeVarDispatch`, which a walk over types alone missed. + +## Notes + +- `%` is real modulo in Wurst; `mod` is integer modulo. `int % 8` types as `real`. +- Emitted Lua must be byte-identical for identical input (AGENTS.md §8). It is the only + emitted output that can be diffed across runs — see item 8. +- Tests run five Jass configurations plus the interpreter, then the Lua target separately. + `testAssertOkLines(true, ...)` covers both the pre-transform interpreter and full + monomorphisation, so it is a stronger check than it looks. +- The stdlib copy under `de.peeeq.wurstscript/temp/WurstStdlib2` is a fetched artefact for + tests. Real stdlib changes belong in the WurstStdlib2 repo, not here. From 6fac1f027073e4225f76280351a3e7dbb65a996d Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 03:48:38 +0200 Subject: [PATCH 03/21] Expand the backlog with a recurrence guard, design blockers and scope --- BACKLOG.md | 59 +++++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 49 insertions(+), 10 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 753288587..286550bec 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -22,36 +22,75 @@ Notes rather than leaving it in a commit message. Overriding methods must keep landing in the same slot, so normalise per distinct original name, not per method node. -2. **`slotFor` looks bound to `get`'s implementation** in the same emitted Lua. May be a real +2. **Nothing checks that emitted Lua identifiers are valid.** The luac syntax check catches a + hard break, but not the silent case: `Class.get_specialized_integer, integer = impl` parses + fine and quietly assigns to two targets. That is how item 1 shipped unnoticed. Assert in the + Lua test harness that every emitted name — variable, function, class, method slot, field — + matches `[A-Za-z_][A-Za-z0-9_]*`, so this whole class of bug fails loudly at the point it is + introduced. Do this alongside item 1; it is what stops item 1 recurring. + +3. **`slotFor` looks bound to `get`'s implementation** in the same emitted Lua. May be a real mis-binding in `specializeMethod`/`adaptSubmethods`, may be an artefact of item 1 mangling the output. Diagnose only after item 1, from freshly emitted Lua. -3. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. +4. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. Add `remove` with tombstones, and an assertion that the emitted code stays cheap: no dispatch node, no instance dictionary, and no WC3 hashtable natives — array access only, which is the whole point versus `HashMap extends Table`. -4. **Lua dispatch inside a closure.** Works on Jass since #1229. On Lua the specialised class +5. **Lua dispatch inside a closure.** Works on Jass since #1229. On Lua the specialised class is built correctly but nothing calls it, because the closure is reached through its interface and `specializeMethod` renames the method out of its dispatch slot. `TypeClassTests.dispatchInsideClosureIsRejectedForLua` pins the current diagnostic and should become a success test. Related to item 1; AGENTS.md flags this machinery. -5. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass. +6. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass. -6. **Module bounds.** `module M` is rejected with a clear message today. Needs +7. **Module bounds.** `module M` is rejected with a clear message today. Needs receiver rewriting during expansion, or type parameters on `ModuleInstanciation`. -7. **One `ImTypeVar` per type parameter.** Name-tolerant lookups remain in +8. **`MOD_INT`/`DIV_INT` return the left operand's type** rather than `int` + (`AttrExprType.java`, the `case MOD_INT` branch), where `caseMathOperation` returns + `WurstTypeInt.instance()` for `+`, `-`, `*`. Only observable if something is a proper + subtype of int, so it may be harmless — establish whether it is reachable, then either fix + it or leave a comment saying why the asymmetry is intended. Small. + +9. **Keep `WURST_LANGUAGE.md` and `CHANGELOG.md` current** as items land. The bounds section + says nothing about closures, which now work on Jass. Fold this into whichever item changes + the behaviour rather than doing it as a separate pass. + +10. **One `ImTypeVar` per type parameter.** Name-tolerant lookups remain in `EliminateGenerics.indexOfTypeVar`, `inheritTypeClassBinding` and `ProgramState.getCurrentTypeArgument`, compensating for several nodes standing for one source parameter. Making the node canonical lets all three compare by identity and removes a class of silent wrong dispatch. Mechanical, well covered by the suite. -8. **Jass temp counter is not reset between compilations.** Two runs of the same commit emit - different `.j` (`temp151` vs `temp8`) because the counter is JVM-wide and depends on how - many tests ran before. Not wrong for compiling one map, but it means `.j` cannot be diffed - across runs to validate a change — only `.lua` can. Fixing it would make Jass diffable. +11. **Jass temp counter is not reset between compilations.** Two runs of the same commit emit + different `.j` (`temp151` vs `temp8`) because the counter is JVM-wide and depends on how + many tests ran before. Not wrong for compiling one map, but it means `.j` cannot be diffed + across runs to validate a change — only `.lua` can. Fixing it would make Jass diffable. + +12. **Standing item, never finished.** When nothing above is left, find the next thing worth + doing and add it here rather than stopping. Good sources, in order: a test that would have + caught a bug already found; a place where two mechanisms do the same job and disagree; a + comment claiming something the code no longer does; a path where a wrong result is silent + rather than loud. Add what is found as a numbered item and start on it. + +## Blocked on a decision + +- **Eliminating the remaining `castTo int`.** The motivating case is timer data attachment + (`ClosureTimers.wurst`), and the containers behind it: `Table` has 81 casts, `HashList` 13, + `HashSet` 6, `HashMap` 4. None can adopt bounds as things stand, because an instance is + declared one type at a time and these accept any type. It needs a way to give an instance + for a whole family — every class type, or every handle type — which is a language design + question: what the syntax is, where such an instance may be declared under the orphan rule, + and whether a specific instance always beats a family one. Do not start this autonomously. + +## Out of scope + +- The stdlib itself. `de.peeeq.wurstscript/temp/WurstStdlib2` is a fetched artefact for tests; + editing it changes nothing real. `FastHashMap` ships from the WurstStdlib2 repo once the + compiler-side proof is complete, and that is a separate decision. ## Done From e4da96d7d5d032d07da28a51fc9be608f6b12bed Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 03:52:22 +0200 Subject: [PATCH 04/21] Add the overnight loop brief --- .claude/ralph-loop.local.md | 10 ++++++++ LOOP.md | 48 +++++++++++++++++++++++++++++++++++++ 2 files changed, 58 insertions(+) create mode 100644 .claude/ralph-loop.local.md create mode 100644 LOOP.md diff --git a/.claude/ralph-loop.local.md b/.claude/ralph-loop.local.md new file mode 100644 index 000000000..305f01377 --- /dev/null +++ b/.claude/ralph-loop.local.md @@ -0,0 +1,10 @@ +--- +active: true +iteration: 1 +session_id: 0413fc86-e9a2-4954-bc47-1620ed81d95e +max_iterations: 0 +completion_promise: null +started_at: "2026-08-15T01:51:05Z" +--- + +Work on the WurstScript compiler at c:Usersrun-eDocumentsGitHubWurstScript, on the branch feat/fasthashmap-proof. Read AGENTS.md and BACKLOG.md before doing anything. diff --git a/LOOP.md b/LOOP.md new file mode 100644 index 000000000..ed3234d69 --- /dev/null +++ b/LOOP.md @@ -0,0 +1,48 @@ +# Overnight loop brief + +Work on the WurstScript compiler in this repository, on the branch +`feat/fasthashmap-proof`. Read `AGENTS.md` and `BACKLOG.md` before doing anything. + +## Each iteration + +1. Read `BACKLOG.md`. Take the top item that isn't blocked. +2. Do it. Root-cause it — adjust the underlying system rather than patching a symptom. + If it turns out to be bigger than one iteration, split it in the backlog and do the + first part. +3. Verify. Targeted tests while iterating; the full suite before any commit that touches + main source: + + cd de.peeeq.wurstscript && ./gradlew test + + It must be green. Emitted Lua must stay byte-identical unless the change is meant to + alter it — compare two runs to check. Do not diff `.j` across runs; it is not stable, + and backlog item 11 explains why. +4. Commit and push. Small commits, one concern each. Never end an iteration with + uncommitted work. +5. Update `BACKLOG.md`: move finished items to Done with one line on what actually + happened, add anything learned to Notes, reorder if something more urgent turned up. + +## Rules + +- Never force-push, rewrite history, merge, push to master, or open a PR. The branch gets + reviewed in the morning. +- Commits are authored as the repository owner. No AI, assistant, or co-author references + anywhere in commit messages or code comments. +- Never stop to ask. If something needs a decision from the owner, write the question into + `BACKLOG.md` under that item, mark it blocked, and move to the next item. +- If the full suite goes red and it can't be fixed within the iteration, revert the change, + note why in `BACKLOG.md`, and move on. Leave the branch green. +- Comments explain why, not what. Match the surrounding style. +- Do not touch `de.peeeq.wurstscript/temp/WurstStdlib2` — a fetched test artefact, editing + it changes nothing real. + +## Keeping going + +- Finishing an item is not the end of the run. Go back to step 1 and take the next one. Do + not stop to summarise, and do not treat a green suite as a finish line. +- At most one item may be marked blocked per iteration, and only if it genuinely needs a + decision rather than more work. If everything left looks blocked, that is wrong about at + least one of them — re-read and start the one that can be moved furthest. +- If an item goes three iterations without landing a commit, split it in the backlog and + move on. Don't spend the whole run on one thing. +- Backlog item 12 is a standing item that never completes. There is always a next item. From cc994da12d2235f0f04fcdd5f7944cc0fdfd3222 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 04:15:00 +0200 Subject: [PATCH 05/21] Sanitise Lua method names where they are assigned Method names become Lua table keys, so they must be identifiers. A method specialised with two type arguments was named after them, commas included, and emitted `Class.get_specialized_integer, integer = impl` - valid Lua that quietly assigns to two targets; a tuple argument produced characters luac rejects outright. normalizeMethodNames is the pass that gives one name to a whole dispatch group, so it sanitises before uniquing: two names that differed only in characters Lua has no place for still get a slot each. The backend maps every slot key and every LuaMethod name through the same function, so call sites and class tables keep agreeing. Lua's identifier rule now has one home. The luac check never caught this, because the broken output parses. Assert instead on the names themselves: every emitted function, method, variable, field and call-by-name must be an identifier, checked for every testLua compile. --- BACKLOG.md | 52 ++++++++------ .../imtranslation/LuaDispatchPreparation.java | 5 +- .../lua/translation/LuaAssertions.java | 70 +++++++++++++++++++ .../lua/translation/LuaIdentifiers.java | 54 ++++++++++++++ .../lua/translation/LuaTranslator.java | 45 +++++------- .../wurstscript/tests/WurstScriptTest.java | 2 + 6 files changed, 177 insertions(+), 51 deletions(-) create mode 100644 de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java diff --git a/BACKLOG.md b/BACKLOG.md index 286550bec..ba7ec03be 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -10,28 +10,22 @@ Notes rather than leaving it in a commit message. ## Todo -1. **Lua method names reach the backend unsanitised.** `LuaTranslator.luaMethod.initFor` - passes `a.getName()` raw; `luaVar`, `luaFunc` and `luaClassVar` all sanitise via - `uniqueName`. Method names become Lua table keys, so they must be valid identifiers. - `EliminateGenerics.specializeMethod` builds `name + "_specialized_" + generics.makeName()` - on the Lua path, and `makeName` joins type arguments with `", "` — so a class method - specialised with two type arguments emits `Class.get_specialized_integer, integer = impl`, - which is valid Lua assigning to two targets, and a tuple argument emits `⦅⦆` and fails - the syntax check. Repro: `FastHashMapTests.tupleKeyLua`. Commas are already visible in - `test-output/lua/FastHashMapTests_fastHashMapRuntimeLua.lua`. - Overriding methods must keep landing in the same slot, so normalise per distinct original - name, not per method node. - -2. **Nothing checks that emitted Lua identifiers are valid.** The luac syntax check catches a - hard break, but not the silent case: `Class.get_specialized_integer, integer = impl` parses - fine and quietly assigns to two targets. That is how item 1 shipped unnoticed. Assert in the - Lua test harness that every emitted name — variable, function, class, method slot, field — - matches `[A-Za-z_][A-Za-z0-9_]*`, so this whole class of bug fails loudly at the point it is - introduced. Do this alongside item 1; it is what stops item 1 recurring. - -3. **`slotFor` looks bound to `get`'s implementation** in the same emitted Lua. May be a real - mis-binding in `specializeMethod`/`adaptSubmethods`, may be an artefact of item 1 mangling - the output. Diagnose only after item 1, from freshly emitted Lua. +Numbering is stable: finished items leave a gap rather than shifting the ones below, +because `LOOP.md` refers to items by number. + +3. **`slotFor` is bound to `get`'s implementation** in the emitted Lua — confirmed real, and + not an artefact of item 1: it survives sanitisation unchanged. Diagnosed, not yet fixed. + The alias sets in `LuaDispatchPreparation` decide which slots a method claims, and + `sharesSemanticName` accepts a match on *either* of two names: the source name (`get`, + `slotFor`) or `semanticNameFromMethodName`, which is the substring after the last + underscore. For a specialised method that substring is a type-argument fragment — + `FastHashMap_get_specialized__integer__integer___integer` and the `slotFor` one both yield + `integer` — so two unrelated methods count as sharing a name. They also share a dispatch + signature here (`(pos) returns int` both), which is the other half of the guard, so `get` + claims `slotFor`'s slot. Fix: when both methods have a real source name, that should decide; + the substring heuristic is a fallback for when there is no trace to ask, not an alternative. + Watch the closure and bridge cases in `TypeClassTests`/`LuaBackendAuditTests` — they are what + the loose match was presumably widened for. 4. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. Add `remove` with tombstones, and an assertion that the emitted code stays cheap: no @@ -94,6 +88,14 @@ Notes rather than leaving it in a commit message. ## Done +- 1 + 2. Lua method names are sanitised where they are assigned, not where they are printed. + `LuaDispatchPreparation.normalizeMethodNames` is the pass that gives one name to a whole + dispatch group, so it now sanitises before uniquing — two names differing only in characters + Lua has no place for still get a slot each. `LuaTranslator` maps every slot key and every + `LuaMethod` name through the same function, so call sites and class tables agree. Lua's + identifier rule now lives in one place, `LuaIdentifiers`. `LuaAssertions.assertNamesAreValidIdentifiers` + walks the emitted Lua and fails on any name that is not an identifier; it runs for every + `testLua` compile, so the silent two-target-assignment case cannot come back. - Substitution now carries the type class binding with the type (#1229). Also fixed the type-variable reference on `ImTypeVarDispatch`, which a walk over types alone missed. @@ -101,7 +103,11 @@ Notes rather than leaving it in a commit message. - `%` is real modulo in Wurst; `mod` is integer modulo. `int % 8` types as `real`. - Emitted Lua must be byte-identical for identical input (AGENTS.md §8). It is the only - emitted output that can be diffed across runs — see item 8. + emitted output that can be diffed across runs — see item 11. +- Method names are not what the frontend called them. `LuaDispatchPreparation` renames a whole + dispatch group to one name and attaches alias sets, and only then does the backend run. A + question about which Lua slot something lands in is a question about that pass, not about + `LuaTranslator`. - Tests run five Jass configurations plus the interpreter, then the Lua target separately. `testAssertOkLines(true, ...)` covers both the pre-transform interpreter and full monomorphisation, so it is a stronger check than it looks. diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java index 662c812b9..cbed8db1b 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java @@ -11,6 +11,7 @@ import de.peeeq.wurstscript.jassIm.ImProg; import de.peeeq.wurstscript.jassIm.ImType; import de.peeeq.wurstscript.jassIm.ImVars; +import de.peeeq.wurstscript.translation.lua.translation.LuaIdentifiers; import java.util.ArrayList; import java.util.Collection; @@ -99,7 +100,9 @@ private static void normalizeMethodNames(ImProg prog, List allMethods) continue; } group.sort(Comparator.comparing(LuaDispatchPreparation::methodSortKey)); - String name = uniqueName(group.get(0).getName(), usedNames); + // The name is about to become a Lua table key. Sanitising before uniquing means two + // names that only differed in characters Lua has no place for still get one slot each. + String name = uniqueName(LuaIdentifiers.toIdentifier(group.get(0).getName()), usedNames); for (ImMethod method : group) { method.setName(name); } diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaAssertions.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaAssertions.java index 82d003d60..4c1ed5298 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaAssertions.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaAssertions.java @@ -1,9 +1,19 @@ package de.peeeq.wurstscript.translation.lua.translation; +import de.peeeq.wurstscript.luaAst.Element; +import de.peeeq.wurstscript.luaAst.LuaCompilationUnit; +import de.peeeq.wurstscript.luaAst.LuaExprFieldAccess; +import de.peeeq.wurstscript.luaAst.LuaExprFunctionCallByName; +import de.peeeq.wurstscript.luaAst.LuaFunction; +import de.peeeq.wurstscript.luaAst.LuaMethod; +import de.peeeq.wurstscript.luaAst.LuaTableNamedField; +import de.peeeq.wurstscript.luaAst.LuaVariable; + import java.util.ArrayList; import java.util.HashSet; import java.util.List; import java.util.Set; +import java.util.TreeSet; /** * Static assertion helpers for the Lua backend. @@ -15,6 +25,66 @@ public class LuaAssertions { private LuaAssertions() {} + /** + * Asserts that every name the backend emits is a Lua identifier. + * + *

A name that is not one usually breaks the syntax check, but not always: a table key + * containing a comma parses as an assignment to two targets and quietly stores the value in + * the wrong place. Checking the names themselves catches that case at the point it is + * introduced, rather than as a wrong result at runtime. + */ + public static void assertNamesAreValidIdentifiers(LuaCompilationUnit luaCode) { + Set invalid = new TreeSet<>(); + luaCode.accept(new Element.DefaultVisitor() { + private void check(String kind, String name) { + // A vararg parameter is the one name that is legal without being an identifier. + if (!LuaIdentifiers.isValid(name) && !LuaIdentifiers.VARARG.equals(name)) { + invalid.add(kind + " '" + name + "'"); + } + } + + @Override + public void visit(LuaFunction f) { + super.visit(f); + check("function", f.getName()); + } + + @Override + public void visit(LuaMethod m) { + super.visit(m); + check("method", m.getName()); + } + + @Override + public void visit(LuaVariable v) { + super.visit(v); + check("variable", v.getName()); + } + + @Override + public void visit(LuaExprFieldAccess fa) { + super.visit(fa); + check("field", fa.getFieldName()); + } + + @Override + public void visit(LuaTableNamedField f) { + super.visit(f); + check("field", f.getFieldName()); + } + + @Override + public void visit(LuaExprFunctionCallByName call) { + super.visit(call); + check("call to", call.getFuncName()); + } + }); + if (!invalid.isEmpty()) { + throw new RuntimeException("Wurst Lua backend assertion failed: emitted names are not Lua identifiers: " + + String.join(", ", invalid)); + } + } + /** * Asserts that every emitted call to {@code __wurst_GetHandleId} has a helper definition. * diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java new file mode 100644 index 000000000..95cde0a8b --- /dev/null +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java @@ -0,0 +1,54 @@ +package de.peeeq.wurstscript.translation.lua.translation; + +/** + * Lua's rule for what a generated name may look like, kept in one place. + * + *

Names from the intermediate language are not constrained to Lua's identifier syntax; + * specialised generics, for example, are named after their type arguments. Sanitising in the + * backend keeps that rule where it belongs rather than requiring every earlier pass to know + * about Lua. Any collisions the mapping introduces are resolved by the usual uniquing. + */ +public final class LuaIdentifiers { + + /** Lua's vararg parameter, which is a legal parameter name but not an identifier. */ + public static final String VARARG = "..."; + + /** Whether {@code name} can be used as-is as a Lua identifier or table key. */ + public static boolean isValid(String name) { + if (name == null || name.isEmpty() || isDigit(name.charAt(0))) { + return false; + } + for (int i = 0; i < name.length(); i++) { + if (!isIdentifierPart(name.charAt(i))) { + return false; + } + } + return true; + } + + /** Maps any name onto a Lua identifier, leaving names that already are one untouched. */ + public static String toIdentifier(String name) { + if (isValid(name)) { + return name; + } + StringBuilder sb = new StringBuilder(name.length()); + for (int i = 0; i < name.length(); i++) { + char c = name.charAt(i); + sb.append(isIdentifierPart(c) ? c : '_'); + } + if (sb.length() == 0 || isDigit(sb.charAt(0))) { + sb.insert(0, '_'); + } + return sb.toString(); + } + + private static boolean isIdentifierPart(char c) { + return c == '_' || (c < 128 && Character.isLetterOrDigit(c)); + } + + private static boolean isDigit(char c) { + return c >= '0' && c <= '9'; + } + + private LuaIdentifiers() {} +} diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaTranslator.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaTranslator.java index 4e729a863..f00242efa 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaTranslator.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaTranslator.java @@ -135,7 +135,9 @@ public LuaFunction initFor(ImFunction a) { @Override public LuaMethod initFor(ImMethod a) { LuaExpr receiver = LuaAst.LuaExprVarAccess(luaClassVar.getFor(a.attrClass())); - return LuaAst.LuaMethod(receiver, a.getName(), LuaAst.LuaParams(), LuaAst.LuaStatements()); + // A method name is a table key, so it must be an identifier - but unlike a variable + // it must not be uniqued: every override has to keep landing in the same slot. + return LuaAst.LuaMethod(receiver, dispatchSlotName(a.getName()), LuaAst.LuaParams(), LuaAst.LuaStatements()); } }; @@ -196,28 +198,8 @@ public LuaTranslator(ImProg prog, ImTranslator imTr) { luaModel = LuaAst.LuaCompilationUnit(); } - /** - * Makes an intermediate-language name usable as a Lua identifier. - *

- * Names from the IM are not constrained to Lua's identifier syntax; specialised generics, for - * example, are named after their type arguments. Sanitising here keeps that rule where it - * belongs, in the backend, rather than requiring every earlier pass to know about Lua. Any - * collisions the mapping introduces are resolved by the usual uniquing. - */ - private static String toLuaIdentifier(String name) { - StringBuilder sb = new StringBuilder(name.length()); - for (int i = 0; i < name.length(); i++) { - char c = name.charAt(i); - sb.append(c == '_' || Character.isLetterOrDigit(c) && c < 128 ? c : '_'); - } - if (sb.length() == 0 || Character.isDigit(sb.charAt(0))) { - sb.insert(0, '_'); - } - return sb.toString(); - } - protected String uniqueName(String rawName) { - String name = toLuaIdentifier(rawName); + String name = LuaIdentifiers.toIdentifier(rawName); Integer nextIndex = uniqueNameCounters.get(name); if (nextIndex == null) { uniqueNameCounters.put(name, 1); @@ -462,7 +444,7 @@ private void collectMethodNames(ImClass c, Set methodNames, Set } visited.add(c); for (ImMethod method : c.getMethods()) { - methodNames.add(method.getName()); + methodNames.add(dispatchSlotName(method.getName())); } for (ImClassType sc : c.getSuperClasses()) { collectMethodNames(sc.getClassDef(), methodNames, visited); @@ -914,7 +896,7 @@ private void createMethods(ImClass c, LuaVariable classVar) { ImMethod chosen = chosenByGroup.get(groupMethods); Set memberNames = new HashSet<>(); for (ImMethod m : groupMethods) { - memberNames.add(m.getName()); + memberNames.add(dispatchSlotName(m.getName())); } Set slotNames = collectDispatchSlotNames(c, groupMethods); for (String slotName : slotNames) { @@ -934,7 +916,7 @@ private void createMethods(ImClass c, LuaVariable classVar) { ImMethod chosen = chosenByGroup.get(groupMethods); Set memberNames = new HashSet<>(); for (ImMethod m : groupMethods) { - memberNames.add(m.getName()); + memberNames.add(dispatchSlotName(m.getName())); } for (String slotName : collectDispatchSlotNames(c, groupMethods)) { if (memberNames.contains(slotName)) { @@ -973,6 +955,15 @@ && implArity(chosen) != implArity(current)) { } } + /** + * The Lua table key a dispatch slot is emitted under. Aliases and class-qualified names are + * built from IM names, which may contain characters Lua has no place for; the mapping has to + * be the same one call sites go through, so that a slot is still found under its new name. + */ + private String dispatchSlotName(String rawName) { + return LuaIdentifiers.toIdentifier(rawName); + } + private Set collectDispatchSlotNames(ImClass receiverClass, List groupMethods) { Set slotNames = new TreeSet<>(); Set semanticNames = new TreeSet<>(); @@ -982,7 +973,7 @@ private Set collectDispatchSlotNames(ImClass receiverClass, List collectDispatchSlotNames(ImClass receiverClass, List()); for (String className : classNames) { for (String semanticName : semanticNames) { - slotNames.add(className + "_" + semanticName); + slotNames.add(dispatchSlotName(className + "_" + semanticName)); } } } diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java index 3498cec3f..545f28c4d 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java @@ -25,6 +25,7 @@ import de.peeeq.wurstscript.jassprinter.JassPrinter; import de.peeeq.wurstscript.luaAst.LuaCompilationUnit; import de.peeeq.wurstscript.luaAst.*; +import de.peeeq.wurstscript.translation.lua.translation.LuaAssertions; import de.peeeq.wurstscript.translation.imtranslation.ImTranslator; import de.peeeq.wurstscript.translation.imtranslation.RecycleCodeGeneratorQueue; import de.peeeq.wurstscript.utils.Utils; @@ -500,6 +501,7 @@ private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, compiler.runCompiletime(WurstProjectConfigData.empty(), false, false); LuaCompilationUnit luaCode = compiler.transformProgToLua(); + LuaAssertions.assertNamesAreValidIdentifiers(luaCode); checkLuaRootPurity(luaCode); StringBuilder sb = new StringBuilder(); luaCode.print(sb, 0); From 596b6fae6ce66d2964cd8e0b41582cb395f3bfe0 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 04:45:31 +0200 Subject: [PATCH 06/21] Bound the wait on the Lua interpreter in tests The execution path read the spawned process's stderr to EOF before touching stdout, and never bounded the wait. A program that fills the stdout pipe blocks writing while the harness blocks reading stderr, and the suite stops with no output, no timeout and no failing test - a stray worker JVM was still sitting on the build directory twenty minutes later. checkLuaSyntax, in the same file, already drained both pipes on their own threads and waited with a timeout. The execution path now uses the same helper, so a program that does not terminate fails its test instead of the run. Also records what building a repro turned up: a bounded generic class cannot be subclassed on either backend, and the div/mod result type asymmetry is reachable through integer literals rather than harmless. --- BACKLOG.md | 50 +++++++++++++++++-- .../wurstscript/tests/WurstScriptTest.java | 32 ++++++------ 2 files changed, 65 insertions(+), 17 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index ba7ec03be..7d6c0ad50 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -45,9 +45,11 @@ because `LOOP.md` refers to items by number. 8. **`MOD_INT`/`DIV_INT` return the left operand's type** rather than `int` (`AttrExprType.java`, the `case MOD_INT` branch), where `caseMathOperation` returns - `WurstTypeInt.instance()` for `+`, `-`, `*`. Only observable if something is a proper - subtype of int, so it may be harmless — establish whether it is reachable, then either fix - it or leave a comment saying why the asymmetry is intended. Small. + `WurstTypeInt.instance()` for `+`, `-`, `*`. It *is* reachable: `WurstTypeIntLiteral` is a + proper subtype of both int and real, and `caseMathOperation` collapses two literals to `int` + precisely so `real r = 1 + 1` stays an error. Returning `leftType` skips that collapse, so + `real r = 7 div 2` and `real r = 7 mod 2` should be accepted where `+` is rejected. Confirm + with a test first — that is the failing repro — then return `WurstTypeInt.instance()`. Small. 9. **Keep `WURST_LANGUAGE.md` and `CHANGELOG.md` current** as items land. The bounds section says nothing about closures, which now work on Jass. Fold this into whichever item changes @@ -64,6 +66,40 @@ because `LOOP.md` refers to items by number. many tests ran before. Not wrong for compiling one map, but it means `.j` cannot be diffed across runs to validate a change — only `.lua` can. Fixing it would make Jass diffable. +13. **A bounded generic class cannot be subclassed.** Found while building a repro for item 3; + both backends break, differently, on the same program: + + interface Show + function show(T x) returns int + implements Show + function show(int x) returns int + return x + class Box + K key + construct(K k) + key = k + function size(int extra) returns int + return K.show(key) + extra + function shift(int extra) returns int + return 1000 + extra + class SubBox extends Box + construct(int k) + super(k) + override function size(int extra) returns int + return super.size(extra) + 100 + init + Box b = new Box(5) + Box s = new SubBox(5) + if b.size(1) == 6 and b.shift(1) == 1001 and s.size(1) == 106 + testSuccess() + + Jass fails to compile: `Typevar dispatch not eliminated.` Lua compiles and runs but never + reaches `testSuccess`: the override makes `size` dispatched, and the emitted call is + `b:Box_size_specialized_integer(1)` while `b` was allocated from the *erased* `Box` table, + which binds only `shift`. The specialised table `Box_specialized_integer` has the slot; the + instance never gets that table. Split this if the two turn out to have separate causes — + the Jass one is loud and probably the smaller of the two. + 12. **Standing item, never finished.** When nothing above is left, find the next thing worth doing and add it here rather than stopping. Good sources, in order: a test that would have caught a bug already found; a place where two mechanisms do the same job and disagree; a @@ -88,6 +124,12 @@ because `LOOP.md` refers to items by number. ## Done +- 14. The Lua execution harness no longer hangs. It read the spawned interpreter's stderr to EOF + before touching stdout, and never bounded the wait, so a program that filled the stdout pipe + deadlocked the whole suite with no output and no timeout — a stray JVM was still sitting on it + twenty minutes later. `checkLuaSyntax`, ten lines further down the same file, already drained + both pipes concurrently and waited with a timeout; the execution path now uses the same + helper. Verified against the item 13 repro: hung indefinitely before, fails in 8s after. - 1 + 2. Lua method names are sanitised where they are assigned, not where they are printed. `LuaDispatchPreparation.normalizeMethodNames` is the pass that gives one name to a whole dispatch group, so it now sanitises before uniquing — two names differing only in characters @@ -104,6 +146,8 @@ because `LOOP.md` refers to items by number. - `%` is real modulo in Wurst; `mod` is integer modulo. `int % 8` types as `real`. - Emitted Lua must be byte-identical for identical input (AGENTS.md §8). It is the only emitted output that can be diffed across runs — see item 11. +- A test that hangs looks exactly like a test that is slow. If the suite stops making progress, + take a thread dump of the forked worker (`jstack `) before killing it — it names the line. - Method names are not what the frontend called them. `LuaDispatchPreparation` renames a whole dispatch group to one name and attaches alias sets, and only then does the backend run. A question about which Lua slot something lands in is a question about that pass, not about diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java index 545f28c4d..da1c2a9e7 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java @@ -48,6 +48,9 @@ public class WurstScriptTest { public static final String TEST_OUTPUT_PATH = "./test-output/"; + + /** Generous enough for the slowest test program, short enough that a loop fails the run. */ + private static final int LUA_EXECUTION_TIMEOUT_SECONDS = 60; private static volatile String resolvedLuaExecutable; private static volatile String resolvedLuacExecutable; private static volatile String extractedLuaWin; @@ -527,7 +530,6 @@ private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, throw new org.testng.SkipException( "Skipped Lua execution (translation and luac syntax check still ran): " + e.getMessage()); } - String line; // Preload the WC3 Lua runtime (Reforged blizzard.j dump + native shim) // when available, so tests execute against real BJ implementations. // The generated script only installs fallbacks for natives that are @@ -550,26 +552,28 @@ private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, Process p = Runtime.getRuntime().exec(args); StringBuilder errors = new StringBuilder(); StringBuilder output = new StringBuilder(); - try (BufferedReader input = new BufferedReader(new InputStreamReader(p.getErrorStream()))) { - while ((line = input.readLine()) != null) { - System.err.println(line); - errors.append(line); - errors.append("\n"); - } + // Both pipes must be drained while the program runs, and the wait must end: a + // generated program that loops forever would otherwise hang the whole suite, + // and one that fills the stdout pipe would deadlock against a stderr-first read. + Thread outCollector = collectStreamAsync(p.getInputStream(), output); + Thread errCollector = collectStreamAsync(p.getErrorStream(), errors); + if (!p.waitFor(LUA_EXECUTION_TIMEOUT_SECONDS, TimeUnit.SECONDS)) { + p.destroyForcibly(); + throw new Error(currentTestEnv + ": Lua program did not terminate within " + + LUA_EXECUTION_TIMEOUT_SECONDS + "s: " + luaFile.getName()); } + outCollector.join(); + errCollector.join(); if (errors.length() > 0) { + System.err.print(errors); throw new TestFailException(errors.toString()); } boolean success = false; - try (BufferedReader input = new BufferedReader(new InputStreamReader(p.getInputStream()))) { - while ((line = input.readLine()) != null) { - if (line.equals("testSuccess")) { - success = true; - } - output.append(line); - output.append("\n"); + for (String outputLine : output.toString().split("\n", -1)) { + if (outputLine.equals("testSuccess")) { + success = true; } } if (!success) { From a12ee4ce4aee376af8472c9664d0ef63814de40f Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 05:00:19 +0200 Subject: [PATCH 07/21] Let declared names decide which dispatch slot a method claims A method claims the slots of every method it might be an override of, and sharesSemanticName accepted a match on either the declared name or the segment after the last underscore of the mangled one. For a specialised method that segment is a fragment of the type argument: get and slotFor of FastHashMap both read as 'integer', and since specialisation gave them the same signature too, get claimed slotFor's slot and the class table bound FastHashMap_slotFor_specialized_integer__integer to FastHashMap_get_specialized. Declared names now settle it whenever both methods have one. The segment stays as the fallback for closures and bridges, which have no declaration to ask - that is the case it was there for. The emitted Lua is the only record of which implementation a slot should hold, so the test asserts it there: every slot named after a method must bind an implementation named after the same method. --- BACKLOG.md | 31 +++++++------ .../imtranslation/LuaDispatchPreparation.java | 43 +++++++++++++++--- .../wurstscript/tests/FastHashMapTests.java | 44 ++++++++++++++++++- 3 files changed, 99 insertions(+), 19 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 7d6c0ad50..620023adb 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,19 +13,14 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. -3. **`slotFor` is bound to `get`'s implementation** in the emitted Lua — confirmed real, and - not an artefact of item 1: it survives sanitisation unchanged. Diagnosed, not yet fixed. - The alias sets in `LuaDispatchPreparation` decide which slots a method claims, and - `sharesSemanticName` accepts a match on *either* of two names: the source name (`get`, - `slotFor`) or `semanticNameFromMethodName`, which is the substring after the last - underscore. For a specialised method that substring is a type-argument fragment — - `FastHashMap_get_specialized__integer__integer___integer` and the `slotFor` one both yield - `integer` — so two unrelated methods count as sharing a name. They also share a dispatch - signature here (`(pos) returns int` both), which is the other half of the guard, so `get` - claims `slotFor`'s slot. Fix: when both methods have a real source name, that should decide; - the substring heuristic is a fallback for when there is no trace to ask, not an alternative. - Watch the closure and bridge cases in `TypeClassTests`/`LuaBackendAuditTests` — they are what - the loose match was presumably widened for. +15. **One junk dispatch slot per specialised class.** Left over from item 3, same heuristic in + the other place it is used. `addDirectAliases` composes `owner.getName() + "_" + + semanticNameFromMethodName(name)`, and for a specialised method that trailing segment is the + type argument, so every method of `FastHashMap` claims the same + `FastHashMap_specialized_integer__integer_integer` slot and the alphabetically first wins. + Nothing calls it, so it is dead weight rather than a wrong result — but it is the same + mistake, and the alias it *should* produce is the class qualified with the declared name. + Fixing it changes emitted slot names, so it wants its own commit and its own suite run. 4. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. Add `remove` with tombstones, and an assertion that the emitted code stays cheap: no @@ -124,6 +119,16 @@ because `LOOP.md` refers to items by number. ## Done +- 3. `slotFor`'s slot no longer holds `get`'s implementation. The alias sets in + `LuaDispatchPreparation` decide which slots a method claims, and `sharesSemanticName` accepted + a match on either the declared name or `semanticNameFromMethodName` — the substring after the + last underscore, which for a specialised method is a fragment of the type argument. Both + `FastHashMap_get_specialized_integer__integer` and the `slotFor` one end in `integer`, and the + two share a dispatch signature, so `get` claimed `slotFor`'s slot. Declared names now settle it + whenever both methods have one; the substring is a fallback for closures and bridges, which + have no declaration to ask. `FastHashMapTests.fastHashMapRuntimeLua` asserts every slot named + after a method binds that method's implementation — it fails on the old code with exactly the + binding above. - 14. The Lua execution harness no longer hangs. It read the spawned interpreter's stderr to EOF before touching stdout, and never bounded the wait, so a program that filled the stdout pipe deadlocked the whole suite with no output and no timeout — a stray JVM was still sitting on it diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java index cbed8db1b..86b95b25c 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/LuaDispatchPreparation.java @@ -176,10 +176,10 @@ private static void addHierarchyAliases(ImMethod method, Set aliases, Ma return; } String dispatchKey = dispatchSignatureKey(method); - collectHierarchyAliases(owner, dispatchKey, semanticNames, aliases, sortedMethodsByClass, new HashSet<>()); + collectHierarchyAliases(owner, method, dispatchKey, semanticNames, aliases, sortedMethodsByClass, new HashSet<>()); } - private static void collectHierarchyAliases(ImClass c, String dispatchKey, Set semanticNames, Set aliases, + private static void collectHierarchyAliases(ImClass c, ImMethod method, String dispatchKey, Set semanticNames, Set aliases, Map> sortedMethodsByClass, Set visited) { if (c == null || !visited.add(c)) { return; @@ -188,7 +188,7 @@ private static void collectHierarchyAliases(ImClass c, String dispatchKey, Set semanticNames(ImMethod method) { return names; } + /** + * Whether {@code candidate} is the same method as {@code method} under a different name, which + * is what makes it worth claiming its slot. + * + *

When both were declared in source, their declared names settle it. The name-derived + * fallback below reads the segment after the last underscore, which for a specialised method + * is a fragment of the type argument: two unrelated methods of one specialisation both end in + * {@code integer} and would otherwise be taken for one another. + */ + private static boolean sharesSemanticName(ImMethod method, ImMethod candidate, Set semanticNames) { + String declared = declaredName(method); + String candidateDeclared = declaredName(candidate); + if (!declared.isEmpty() && !candidateDeclared.isEmpty()) { + return declared.equals(candidateDeclared); + } + return sharesSemanticName(candidate, semanticNames); + } + private static boolean sharesSemanticName(ImMethod method, Set semanticNames) { if (semanticNames.isEmpty()) { return false; @@ -255,6 +273,21 @@ private static boolean sharesSemanticName(ImMethod method, Set semanticN || semanticNames.contains(sourceSemanticName(method)); } + /** The name the method was written with, or empty when there is no declaration to ask. */ + private static String declaredName(ImMethod method) { + if (method == null) { + return ""; + } + de.peeeq.wurstscript.ast.Element trace = method.attrTrace(); + if (trace instanceof FuncDef funcDef) { + return funcDef.getName(); + } + if (trace instanceof AstElementWithFuncName withFuncName) { + return withFuncName.getFuncNameId().getName(); + } + return ""; + } + private static List sortedMethodsForClass(ImClass c, Map> sortedMethodsByClass) { return sortedMethodsByClass.computeIfAbsent(c, key -> { List methods = new ArrayList<>(key.getMethods()); diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java index 50becb1f3..2c8ea988d 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java @@ -1,7 +1,14 @@ package tests.wurstscript.tests; +import com.google.common.base.Charsets; +import com.google.common.io.Files; import org.testng.annotations.Test; +import java.io.File; +import java.io.IOException; +import java.util.regex.Matcher; +import java.util.regex.Pattern; + /** * Exercises type class bounds through the container they were added for: a hash map whose key type * is bounded by {@code Hashable}, so it can hash and compare keys without erasing them to int. @@ -101,8 +108,43 @@ public void fastHashMapRuntime() { } @Test - public void fastHashMapRuntimeLua() { + public void fastHashMapRuntimeLua() throws IOException { test().testLua(true).executeProg().lines(program(fastHashMap(), INT_INSTANCE, USE_WITH_COLLISION)); + assertEachSlotBindsItsOwnMethod(compiledLua("fastHashMapRuntimeLua")); + } + + private static final String[] METHOD_NAMES = {"slotFor", "put", "get", "has", "size"}; + + private String compiledLua(String testName) throws IOException { + return Files.toString(new File("test-output/lua/FastHashMapTests_" + testName + ".lua"), Charsets.UTF_8); + } + + /** + * A dispatch slot is named after the method it belongs to, and nothing else in the emitted Lua + * records which implementation belongs there — so this is the only place the two can be checked + * against each other. Specialisation names every method of one instantiation after the same type + * argument, which is exactly when they are easiest to confuse. + */ + private void assertEachSlotBindsItsOwnMethod(String lua) { + Matcher assignment = Pattern.compile("(?m)^\\s*\\w+\\.(\\w+) = (\\w+)\\s*$").matcher(lua); + while (assignment.find()) { + String slot = assignment.group(1); + String implementation = assignment.group(2); + for (String method : METHOD_NAMES) { + if (namesMethod(slot, method) && !namesMethod(implementation, method)) { + throw new AssertionError("slot '" + slot + "' is named after " + method + + " but binds '" + implementation + "'"); + } + } + } + } + + /** Whether {@code name} carries {@code method} as one of its underscore-separated segments. */ + private static boolean namesMethod(String name, String method) { + return name.equals(method) + || name.startsWith(method + "_") + || name.endsWith("_" + method) + || name.contains("_" + method + "_"); } /** From 5c234916bc16ca44771c22a81794782f7c9f5c6e Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 05:21:33 +0200 Subject: [PATCH 08/21] Finish the FastHashMap proof with remove and a cost assertion remove leaves a tombstone rather than an empty slot, because a probe that stopped at one would miss keys put down beyond it. slotFor passes tombstones over when searching and returns the first one when putting, so removing and re-putting a key reuses its slot instead of lengthening the run. The probe is bounded by capacity, so a table full of tombstones cannot spin. The cost claim is asserted on the least optimised configuration, because it has to hold by construction rather than because the inliner removed it: storage is four plain Jass arrays, no hashtable native is reached for, slotFor takes nothing beyond the receiver and the key, and both requirements are direct calls rather than ExecuteFunc or a dispatch wrapper. With optimisation on, hash(key) becomes key and equals(a, key) becomes a != key. The dispatch_ functions in the output are the nullpointer check every class method gets, not type class dispatch, so the test looks at the real function underneath rather than asserting they are absent. Building the fixture also turned up a silent wrong result in the interpreter, recorded as item 16: a never-written array of a type parameter compares unequal to the type argument's default. --- BACKLOG.md | 36 ++++- .../wurstscript/tests/FastHashMapTests.java | 133 +++++++++++++++++- 2 files changed, 159 insertions(+), 10 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 620023adb..1d7a5d24c 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,6 +13,27 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. +16. **A never-written array of a type parameter reads as nothing, silently.** In the interpreter + only — Jass and Lua both give the type argument's default. `DefaultValue.get(ImTypeVarRef)` + returns `ILconstUnsafeDefault`, whose `isEqualTo` matches only another `ILconstUnsafeDefault`, + so a comparison against the real default is quietly false rather than an error. Repro: + + class Box + private static T array none + static function first() returns T + return none[0] + init + if Box.first() == 0 + testSuccess() + + Passes on every Jass configuration and fails on the pre-transform interpreter run. The plain + `int array` version passes, so this is specific to the type parameter. The interpreter knows + the current type argument (`ProgramState.resolveType`), but `DefaultValue` is a static + attribute with no access to it, and the array's default supplier is bound when the array is + allocated rather than when it is read. Either resolve at read time where the state is in hand, + or make the placeholder throw when used — what it must not do is compare unequal in silence. + Found by `FastHashMapTests`: the tombstone fixture needs a "no value" for `V`. + 15. **One junk dispatch slot per specialised class.** Left over from item 3, same heuristic in the other place it is used. `addDirectAliases` composes `owner.getName() + "_" + semanticNameFromMethodName(name)`, and for a specialised method that trailing segment is the @@ -22,11 +43,6 @@ because `LOOP.md` refers to items by number. mistake, and the alias it *should* produce is the class qualified with the declared name. Fixing it changes emitted slot names, so it wants its own commit and its own suite run. -4. **Finish the FastHashMap proof.** `FastHashMapTests` is the first real use of bounds. - Add `remove` with tombstones, and an assertion that the emitted code stays cheap: no - dispatch node, no instance dictionary, and no WC3 hashtable natives — array access only, - which is the whole point versus `HashMap extends Table`. - 5. **Lua dispatch inside a closure.** Works on Jass since #1229. On Lua the specialised class is built correctly but nothing calls it, because the closure is reached through its interface and `specializeMethod` renames the method out of its dispatch slot. @@ -119,6 +135,16 @@ because `LOOP.md` refers to items by number. ## Done +- 4. The FastHashMap proof is complete. `remove` leaves a tombstone, which `slotFor` passes over + when searching and reuses when putting; the probe is bounded by capacity rather than running + until it finds a gap, so a table full of tombstones cannot spin. `emittedCodeCostsNothingExtra` + asserts the cost claim on the *least* optimised configuration, because it has to hold by + construction rather than by inlining: storage is four plain Jass arrays, no WC3 hashtable native + is reached for, `slotFor` takes nothing beyond the receiver and the key — no instance is threaded + through at runtime — and both requirements are direct calls, not `ExecuteFunc`, not a dispatch + wrapper. In the optimised output `hash(key)` becomes `key` and `equals(a, key)` becomes `a != key`. + The `dispatch_` functions that remain are the nullpointer check every Wurst class method gets, + not type class dispatch; the test is careful to look at the real function under that wrapper. - 3. `slotFor`'s slot no longer holds `get`'s implementation. The alias sets in `LuaDispatchPreparation` decide which slots a method claims, and `sharesSemanticName` accepted a match on either the declared name or `semanticNameFromMethodName` — the substring after the diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java index 2c8ea988d..4b4938434 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java @@ -35,30 +35,65 @@ private static String[] fastHashMap(String... extra) { " private static K array keys", " private static V array values", " private static boolean array used", + // A removed slot cannot go back to empty: a probe that stopped there would miss keys + // put down beyond it. It becomes a tombstone instead - passed over when searching, + // reused when putting. + " private static boolean array dead", + // Never written, so a read yields V's default. That is the only way to say "no value" + // for a type parameter, and it costs an array read rather than a branch. + " private static V array none", " private static int nextFree = 0", " private int base", " private int count = 0", " construct()", " base = nextFree", " nextFree += CAPACITY", + // The slot holding key, or the one it belongs in: the first tombstone passed over, + // else the empty slot the probe stopped at. Capacity is fixed, so a full table + // returns -1 rather than probing forever. " private function slotFor(K key) returns int", " var i = K.hash(key) mod CAPACITY", " if i < 0", " i += CAPACITY", - " while used[base + i] and not K.equals(keys[base + i], key)", + " var firstDead = -1", + " var probes = 0", + " while probes < CAPACITY", + " let s = base + i", + " if used[s] and K.equals(keys[s], key)", + " return s", + " if not used[s] and not dead[s]", + " if firstDead >= 0", + " return firstDead", + " return s", + " if dead[s] and firstDead < 0", + " firstDead = s", " i = (i + 1) mod CAPACITY", - " return base + i", + " probes++", + " return firstDead", " function put(K key, V value)", " let s = slotFor(key)", " if not used[s]", " used[s] = true", + " dead[s] = false", " keys[s] = key", " count++", " values[s] = value", " function get(K key) returns V", - " return values[slotFor(key)]", + " let s = slotFor(key)", + " if s < base or not used[s]", + " return none[0]", + " return values[s]", " function has(K key) returns boolean", - " return used[slotFor(key)]", + " let s = slotFor(key)", + " return s >= base and used[s]", + " function remove(K key) returns boolean", + " let s = slotFor(key)", + " if s < base or not used[s]", + " return false", + " used[s] = false", + " dead[s] = true", + " count--", + " return true", " function size() returns int", " return count", }; @@ -113,7 +148,7 @@ public void fastHashMapRuntimeLua() throws IOException { assertEachSlotBindsItsOwnMethod(compiledLua("fastHashMapRuntimeLua")); } - private static final String[] METHOD_NAMES = {"slotFor", "put", "get", "has", "size"}; + private static final String[] METHOD_NAMES = {"slotFor", "put", "get", "has", "size", "remove"}; private String compiledLua(String testName) throws IOException { return Files.toString(new File("test-output/lua/FastHashMapTests_" + testName + ".lua"), Charsets.UTF_8); @@ -147,6 +182,35 @@ private static boolean namesMethod(String name, String method) { || name.contains("_" + method + "_"); } + /** + * Keys 1, 9 and 2 form one probe run at capacity 8. Removing the middle of it is the case + * tombstones exist for: with the slot marked empty instead, the probe for 2 would stop at it + * and report the key missing. + */ + private static final String[] USE_WITH_REMOVE = { + "init", + " let m = new FastHashMap()", + " m.put(1, 10)", + " m.put(9, 90)", + " m.put(2, 20)", + " if m.remove(9) and not m.remove(9)", + " if m.get(2) == 20 and m.has(2) and not m.has(9) and m.size() == 2", + // The tombstone is reused rather than leaked, so the run stays the same length. + " m.put(9, 91)", + " if m.get(9) == 91 and m.size() == 3 and m.get(1) == 10 and m.get(2) == 20", + " testSuccess()", + }; + + @Test + public void removeLeavesATombstone() { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, USE_WITH_REMOVE)); + } + + @Test + public void removeLeavesATombstoneLua() { + test().testLua(true).executeProg().lines(program(fastHashMap(), INT_INSTANCE, USE_WITH_REMOVE)); + } + /** * A tuple key, which is the case old generics cannot serve at all: a tuple has no int * representation to cast to, so the only way to key a map by one is to say how it hashes. @@ -246,4 +310,63 @@ public void twoInstancesOfOneSpecialisation() { " testSuccess()" })); } + + /** + * The point of bounds over {@code HashMap extends Table}: the requirements are resolved when the + * map is specialised, not carried to runtime. This asserts on the least optimised configuration + * on purpose — the cost has to be absent by construction, not removed afterwards by the inliner. + */ + @Test + public void emittedCodeCostsNothingExtra() throws IOException { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, USE_WITH_COLLISION)); + String jass = compiledJass("emittedCodeCostsNothingExtra_no_opts"); + + // Storage is plain Jass arrays, so a lookup is an array read, not a native call. + for (String[] storage : new String[][]{{"integer", "keys"}, {"integer", "values"}, + {"boolean", "used"}, {"boolean", "dead"}}) { + assertMatches(jass, "(?m)^" + storage[0] + " array FastHashMap_" + storage[1] + "\\w*$", + "storage array " + storage[1]); + } + for (String hashtableNative : new String[]{"InitHashtable", "SaveInteger", "LoadInteger", + "SaveStr", "LoadStr", "FlushChildHashtable", "GetHandleId"}) { + if (jass.contains(hashtableNative)) { + throw new AssertionError("the map reached for a WC3 hashtable native: " + hashtableNative); + } + } + + // The bound resolves to a direct call to the instance's own function. An instance passed at + // runtime would show up as an extra parameter here, and a dispatched one as an indirection. + Matcher slotFor = Pattern + // Not the dispatch_ wrapper of the same name - that one is the nullpointer check every + // class method gets, and it is the real function underneath that has to be free of cost. + .compile("(?m)^function (?!dispatch_)\\w*slotFor\\w* takes ([^\\n]*?) returns [^\\n]*\\n(.*?)\\nendfunction", + Pattern.DOTALL) + .matcher(jass); + if (!slotFor.find()) { + throw new AssertionError("no slotFor in the emitted Jass:\n" + jass); + } + if (!slotFor.group(1).equals("integer this, integer key")) { + throw new AssertionError("slotFor carries something beyond the receiver and the key: " + + slotFor.group(1)); + } + String body = slotFor.group(2); + assertMatches(body, "(? Date: Sat, 15 Aug 2026 05:51:22 +0200 Subject: [PATCH 09/21] Take a closure's instantiation from its construction on Lua Lua specialises what it can reach from a concrete type, and it finds that type on calls that carry type arguments. A closure has no such call: it is reached through the interface it implements, which is not generic, so nothing at the call site says what the body dispatches on. Only the construction knows. Three pieces were missing, each of which already exists for the Jass path: the instantiation is now collected from the allocation, the member access is collected so the write that captures the environment lands on the specialised field rather than being dropped as a dead store, and the specialised methods are bound to the roots their originals were submethods of. Registering the original implementation as specialised lets the dispatch left in it settle rather than reaching a backend with no way to express it. All three are gated on the class being closure-generated, which is the case where no call names the instantiation. Ungated, the object comes from the specialised class while its methods were bound to the erased one, and every FastHashMap Lua test fails - the disagreement AGENTS.md section 9 describes. dispatchInsideClosureIsRejectedForLua pinned the old diagnostic and becomes dispatchInsideClosureLua. The language doc and changelog said Jass only. --- BACKLOG.md | 25 +++- CHANGELOG.md | 5 +- .../imtranslation/EliminateGenerics.java | 124 ++++++++++++++++++ .../resources/agent-docs/WURST_LANGUAGE.md | 8 ++ .../wurstscript/tests/TypeClassTests.java | 13 +- 5 files changed, 160 insertions(+), 15 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 1d7a5d24c..a8b941445 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,6 +13,12 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. +17. **A failing Lua test reports a Jass configuration's name.** `translateAndTestLua` never calls + `setCurrentTestEnv`, so it throws with whatever the last Jass configuration left in the field: + a Lua failure reads `With Inlining, Optimizations and Stacktraces: Succeed function not + called`. That is not a small thing to read at 3am — it sent this run looking for a Jass + regression that did not exist. One line, next to the other harness fix. Small. + 16. **A never-written array of a type parameter reads as nothing, silently.** In the interpreter only — Jass and Lua both give the type argument's default. `DefaultValue.get(ImTypeVarRef)` returns `ILconstUnsafeDefault`, whose `isEqualTo` matches only another `ILconstUnsafeDefault`, @@ -43,12 +49,6 @@ because `LOOP.md` refers to items by number. mistake, and the alias it *should* produce is the class qualified with the declared name. Fixing it changes emitted slot names, so it wants its own commit and its own suite run. -5. **Lua dispatch inside a closure.** Works on Jass since #1229. On Lua the specialised class - is built correctly but nothing calls it, because the closure is reached through its - interface and `specializeMethod` renames the method out of its dispatch slot. - `TypeClassTests.dispatchInsideClosureIsRejectedForLua` pins the current diagnostic and - should become a success test. Related to item 1; AGENTS.md flags this machinery. - 6. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass. 7. **Module bounds.** `module M` is rejected with a clear message today. Needs @@ -135,6 +135,19 @@ because `LOOP.md` refers to items by number. ## Done +- 5 (+ the part of 9 that follows it). A type class bound now dispatches from inside a closure on + Lua, and `TypeClassTests.dispatchInsideClosureLua` is a success test. The note in this file was + wrong about the cause: no specialised class was being built at all. Lua specialisation is driven + by calls that carry type arguments, and a closure has none — it is reached through the interface + it implements, which is not generic, so only the construction knows the instantiation. Three + pieces were missing, all present already for Jass: collect the instantiation from `ImAlloc`, + collect the member access so the capture write lands on the specialised field, and bind the + specialised methods to the roots the originals were submethods of (registering the original + implementation as specialised so `settleRemainingDispatches` neutralises what it leaves behind). + All three are gated on the class being closure-generated. Widening them to any constructed class + made the two mechanisms disagree — the object came from the specialised class while its methods + were bound to the erased one — and broke every FastHashMap Lua test, which is the shape of + regression AGENTS.md §9 warns about. `WURST_LANGUAGE.md` and `CHANGELOG.md` say so now. - 4. The FastHashMap proof is complete. `remove` leaves a tombstone, which `slotFor` passes over when searching and reuses when putting; the probe is bounded by capacity rather than running until it finds a gap, so a table full of tombstones cannot spin. `emittedCodeCostsNothingExtra` diff --git a/CHANGELOG.md b/CHANGELOG.md index 413b51211..ededc734f 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -37,8 +37,9 @@ return () -> T.toIndex(x) Substituting a type variable now carries the instance chosen for it along with the type, rather than the - type alone, so lifting a body into a class of its own no longer loses it. Jass only for now: Lua reaches - such a class through its interface and still reports the bound as unresolvable there. + type alone, so lifting a body into a class of its own no longer loses it. This works on both targets. + Lua reaches such a class through the interface it implements, so no call names the instantiation and the + construction is what the specialisation is taken from. - Added new pseudo-natives for debugging memory leaks: diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java index 6baf40263..18d0a83fe 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java @@ -5,6 +5,7 @@ import de.peeeq.wurstscript.WLogger; import de.peeeq.wurstscript.ast.ClassDef; import de.peeeq.wurstscript.ast.ConstructorDef; +import de.peeeq.wurstscript.ast.ExprClosure; import de.peeeq.wurstscript.ast.InterfaceDef; import de.peeeq.wurstscript.ast.PackageOrGlobal; import de.peeeq.wurstscript.ast.WPackage; @@ -188,6 +189,18 @@ public void visit(ImMethodCall call) { super.visit(call); collectGenericNewUse(call); } + + @Override + public void visit(ImAlloc alloc) { + super.visit(alloc); + collectGenericNewUse(alloc); + } + + @Override + public void visit(ImMemberAccess memberAccess) { + super.visit(memberAccess); + collectGenericNewUse(memberAccess); + } }); } @@ -204,6 +217,18 @@ public void visit(ImMethodCall call) { super.visit(call); collectGenericNewUse(call); } + + @Override + public void visit(ImAlloc alloc) { + super.visit(alloc); + collectGenericNewUse(alloc); + } + + @Override + public void visit(ImMemberAccess memberAccess) { + super.visit(memberAccess); + collectGenericNewUse(memberAccess); + } }); } @@ -225,6 +250,61 @@ && functionNeedsSpecialization(call.getFunc(), Collections.newSetFromMap(new Ide } } + /** + * A construction states an instantiation that no call site has to mention. A closure is the case + * that needs it: its class is built from the enclosing type variables and reached through its + * interface, so the call carries no type arguments at all and only the allocation knows what the + * body dispatches on. Restricted to classes that actually dispatch on a bound, so this stays a + * targeted specialisation rather than general monomorphisation on Lua. + */ + private void collectGenericNewUse(ImAlloc alloc) { + ImClassType clazz = alloc.getClazz(); + if (clazz.getTypeArguments().isEmpty() + || typeArgumentsContainTypeVariable(clazz.getTypeArguments()) + || !isConstructionOnlyInstantiation(clazz.getClassDef())) { + return; + } + genericsUses.add(new GenericClazzUse(alloc)); + } + + /** + * Whether the construction is the only place a class's instantiation is stated. + *

+ * A class the user writes is used through calls that carry its type arguments, and those already + * specialise what they need onto the erased class. A closure has no such call: it is reached + * through the interface it implements, which is not generic, so the allocation is the only thing + * that knows what the body dispatches on. Widening this beyond that case makes the two + * mechanisms disagree — the object comes from the specialised class while its methods were bound + * to the erased one. + */ + private boolean isConstructionOnlyInstantiation(ImClass classDef) { + return classDef.attrTrace() instanceof ExprClosure && classNeedsSpecialization(classDef); + } + + /** + * A field of a class specialised from a construction has to be reached on the copy. The write + * that captures a closure's environment is the case that needs it: it names the field of the + * generic class, which nothing allocates any more once the construction was redirected. + */ + private void collectGenericNewUse(ImMemberAccess memberAccess) { + ImVar field = memberAccess.getVar(); + if (field.getParent() == null || !(field.getParent().getParent() instanceof ImClass owningClass)) { + return; + } + if (!isConstructionOnlyInstantiation(owningClass)) { + return; + } + if (memberAccess.getTypeArguments().isEmpty()) { + // The access names a field, not an instantiation; the receiver is what knows which one. + addMemberTypeArguments(memberAccess, owningClass); + } + if (memberAccess.getTypeArguments().isEmpty() + || typeArgumentsContainTypeVariable(memberAccess.getTypeArguments())) { + return; + } + genericsUses.add(new GenericMemberAccess(memberAccess)); + } + private void collectGenericNewUse(ImMethodCall call) { if (specializedCallSites.contains(call)) { return; @@ -1316,12 +1396,56 @@ private ImClass specializeClass(ImClass c, GenericTypes generics) { // NEW: Create specialized global variables for this class instantiation createSpecializedGlobals(c, generics, typeVars); + if (genericNewOnly && isConstructionOnlyInstantiation(c)) { + attachSpecializedClassMethods(c, newC, generics); + } onSpecializedClassTriggers.get(c).forEach(consumer -> consumer.accept(generics, newC)); return newC; } + /** + * Makes the methods of a class specialised from a construction reachable. + *

+ * A class specialised because a call named its instantiation is reached through that call. + * One specialised because it was constructed is not: the receiver is held as its interface, so + * dispatch goes through the root method, whose submethods still list only the generic original. + * Each copy is bound to the same roots, and the original's implementation is recorded as having + * a specialisation so the dispatch left behind in it settles instead of reaching the backend. + */ + private void attachSpecializedClassMethods(ImClass original, ImClass specialized, GenericTypes generics) { + List originalMethods = original.getMethods(); + List specializedMethods = specialized.getMethods(); + if (originalMethods.size() != specializedMethods.size()) { + // The copy is structural, so this cannot happen; bail rather than pair the wrong ones. + return; + } + Map specializationOf = new IdentityHashMap<>(); + for (int i = 0; i < originalMethods.size(); i++) { + ImMethod copy = specializedMethods.get(i); + copy.setMethodClass(JassIm.ImClassType(specialized, JassIm.ImTypeArguments())); + specializationOf.put(originalMethods.get(i), copy); + + ImFunction implementation = originalMethods.get(i).getImplementation(); + ImFunction copyImplementation = copy.getImplementation(); + if (implementation != null && copyImplementation != null && implementation != copyImplementation + && specializedFunctions.get(implementation, generics) == null) { + specializedFunctions.put(implementation, generics, copyImplementation); + } + } + for (ImClass c : new ArrayList<>(prog.getClasses())) { + for (ImMethod root : c.getMethods()) { + for (ImMethod sub : new ArrayList<>(root.getSubMethods())) { + ImMethod copy = specializationOf.get(sub); + if (copy != null && !root.getSubMethods().contains(copy)) { + root.getSubMethods().add(copy); + } + } + } + } + } + private ImExpr rewriteGenericGlobalsInExpr(ImExpr e, ImClass owningClass, GenericTypes generics) { e.accept(new Element.DefaultVisitor() { @Override public void visit(ImVarAccess va) { diff --git a/de.peeeq.wurstscript/src/main/resources/agent-docs/WURST_LANGUAGE.md b/de.peeeq.wurstscript/src/main/resources/agent-docs/WURST_LANGUAGE.md index 3b8b2f538..c094ba98f 100644 --- a/de.peeeq.wurstscript/src/main/resources/agent-docs/WURST_LANGUAGE.md +++ b/de.peeeq.wurstscript/src/main/resources/agent-docs/WURST_LANGUAGE.md @@ -186,6 +186,14 @@ Instances are unique and must be declared next to what they relate. An instance An instance must implement each requirement with the signature it has after the interface's type parameter is replaced by the instance type; a matching name is not enough. An interface used as a bound must not extend another interface, because the requirements of a bound are the interface's own functions. +A requirement can also be dispatched from inside a closure written in a bounded generic. The closure captures the type parameter along with the values it uses, so the instance is still chosen by the caller's type argument: + +```wurst +function foo(Q x) returns int + Producer p = () -> Q.toIndex(x) + return p.produce() +``` + A generic which passes its own type parameter to another bounded generic must declare that bound itself: ```wurst diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index 6e0ba186c..7013be1ce 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -94,15 +94,14 @@ public void dispatchInsideClosure() { } /** - * The same closure is still rejected for Lua, and this pins that it is rejected clearly rather - * than mistranslated. Lua keeps generics erased and specialises only what it can reach through a - * concrete type; a closure is reached through its interface, so the specialised class exists but - * nothing calls it. Making that work is a change to Lua's erasure, not to substitution. Should - * it be made, this test fails and becomes the success case above. + * The same closure on Lua, which keeps generics erased and specialises only what it can reach + * from a concrete type. A closure is reached through its interface, so no call names the + * instantiation — the construction is the only thing that knows it, and that is what the + * specialisation is now driven from. */ @Test - public void dispatchInsideClosureIsRejectedForLua() { - test().testLua(true).executeProg().expectError("could not be resolved for the Lua target").lines( + public void dispatchInsideClosureLua() { + test().testLua(true).executeProg().lines( "package test", "native testSuccess()", "interface ToIndex", From 00f89a609d63ece72a002818085a7cd22fff342e Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 06:07:36 +0200 Subject: [PATCH 10/21] Pin the constructor case for Lua and say what it needs A bound dispatched from a constructor works on Jass and is rejected on Lua. dispatchInsideConstructor covers the working target, and dispatchInsideConstructorIsRejectedForLua pins that the rejection is a clear diagnostic rather than a mistranslation. Not the same gap as the closure case, and the fix from it does not reach. A constructor belongs to its class rather than to a generic function of its own, so the call that runs it carries no type arguments at all: the intermediate language has b = new_Box(21) with b typed Box and new_Box still generic. The calls inside it do carry the class's type variable, but nothing gives the outermost one a concrete argument, and collection requires non-empty type arguments to start. The backlog records where the instantiation could come from instead. --- BACKLOG.md | 50 ++++++++++++------- .../wurstscript/tests/TypeClassTests.java | 41 +++++++++++++++ 2 files changed, 73 insertions(+), 18 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index a8b941445..04e4e650c 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,12 +13,6 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. -17. **A failing Lua test reports a Jass configuration's name.** `translateAndTestLua` never calls - `setCurrentTestEnv`, so it throws with whatever the last Jass configuration left in the field: - a Lua failure reads `With Inlining, Optimizations and Stacktraces: Succeed function not - called`. That is not a small thing to read at 3am — it sent this run looking for a Jass - regression that did not exist. One line, next to the other harness fix. Small. - 16. **A never-written array of a type parameter reads as nothing, silently.** In the interpreter only — Jass and Lua both give the type argument's default. `DefaultValue.get(ImTypeVarRef)` returns `ILconstUnsafeDefault`, whose `isEqualTo` matches only another `ILconstUnsafeDefault`, @@ -49,22 +43,35 @@ because `LOOP.md` refers to items by number. mistake, and the alias it *should* produce is the class qualified with the declared name. Fixing it changes emitted slot names, so it wants its own commit and its own suite run. -6. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass. +6. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass; there is now + a repro for both targets, `TypeClassTests.dispatchInsideConstructor` and + `dispatchInsideConstructorIsRejectedForLua`, the second pinning the current diagnostic. + + Not the same gap as item 5, and the fix from it does not reach: a constructor belongs to the + class rather than to a generic function of its own, so the call that runs it carries no type + arguments at all. The intermediate language has `b = new_Box(21)` with `b` typed + `Box`, and `new_Box` still generic; the calls *inside* it + (`construct_Box`, `Box_init`) do carry the class's type variable, but nothing gives the + outermost one a concrete argument. `collectGenericNewUse` requires non-empty type arguments, so + it never starts. + + The instantiation is only on the type of what the call is assigned to. Three ways to get at it, + roughly in order of how much they would disturb: attach the class's type arguments to + constructor calls when the intermediate language is built, which is where the frontend still + knows them and would serve both targets uniformly — but it changes the Jass path, which reaches + the same answer another way today, so the emitted `.j` needs checking; read them from the + assignment target on the Lua path, which is a syntactic shape and would miss + `foo(new Box(21))`; or specialise from the `#alloc` inside the constructor, which is the + item 5 mechanism but would have to reach back out to the caller. The first looks right; confirm + it is what the Jass path already relies on before changing it. 7. **Module bounds.** `module M` is rejected with a clear message today. Needs receiver rewriting during expansion, or type parameters on `ModuleInstanciation`. -8. **`MOD_INT`/`DIV_INT` return the left operand's type** rather than `int` - (`AttrExprType.java`, the `case MOD_INT` branch), where `caseMathOperation` returns - `WurstTypeInt.instance()` for `+`, `-`, `*`. It *is* reachable: `WurstTypeIntLiteral` is a - proper subtype of both int and real, and `caseMathOperation` collapses two literals to `int` - precisely so `real r = 1 + 1` stays an error. Returning `leftType` skips that collapse, so - `real r = 7 div 2` and `real r = 7 mod 2` should be accepted where `+` is rejected. Confirm - with a test first — that is the failing repro — then return `WurstTypeInt.instance()`. Small. - -9. **Keep `WURST_LANGUAGE.md` and `CHANGELOG.md` current** as items land. The bounds section - says nothing about closures, which now work on Jass. Fold this into whichever item changes - the behaviour rather than doing it as a separate pass. +9. **Keep `WURST_LANGUAGE.md` and `CHANGELOG.md` current** as items land — a standing practice + rather than a task to finish. Fold it into whichever item changes the behaviour rather than + doing it as a separate pass. Both now cover closures on either target, which is what this item + originally pointed at. 10. **One `ImTypeVar` per type parameter.** Name-tolerant lookups remain in `EliminateGenerics.indexOfTypeVar`, `inheritTypeClassBinding` and @@ -135,6 +142,13 @@ because `LOOP.md` refers to items by number. ## Done +- 8. `div` and `mod` return int rather than the left operand's type, matching `caseMathOperation`. + Reachable, not harmless: an integer literal is a proper subtype of both int and real, and + addition collapses two of them to int precisely so `real r = 1 + 1` stays an error — returning + `leftType` skipped that, so `real r = 7 div 2` was accepted. Three tests in `ExpressionTests`: + both operators rejected against a real, and both still int. +- 17. A failing Lua test says so. `translateAndTestLua` now sets the environment label instead of + reporting under whatever Jass configuration ran last. - 5 (+ the part of 9 that follows it). A type class bound now dispatches from inside a closure on Lua, and `TypeClassTests.dispatchInsideClosureLua` is a success test. The note in this file was wrong about the cause: no specialised class was being built at all. Lua specialisation is driven diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index 7013be1ce..d077d2707 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -120,6 +120,47 @@ public void dispatchInsideClosureLua() { ); } + /** + * A constructor runs before the object exists, so the bound has to be resolved from the type + * argument the construction names rather than from anything reachable on the receiver. + */ + private static final String[] DISPATCH_IN_CONSTRUCTOR = { + "package test", + "native testSuccess()", + "interface Show", + " function show(T x) returns int", + "implements Show", + " function show(int x) returns int", + " return x * 2", + "class Box", + " int cached", + " construct(T x)", + " cached = T.show(x)", + "init", + " let b = new Box(21)", + " if b.cached == 42", + " testSuccess()", + }; + + @Test + public void dispatchInsideConstructor() { + testAssertOkLines(true, DISPATCH_IN_CONSTRUCTOR); + } + + /** + * Still rejected for Lua, and this pins that it is rejected clearly rather than mistranslated. + * A constructor belongs to the class, not to a generic function of its own, so the call that + * runs it carries no type arguments — {@code new_Box(21)} in the intermediate language, with + * the instantiation only on the type of what it is assigned to. Nothing on the Lua path reads + * it from there, so the dispatch inside the constructor is never given a concrete type. + * Should that be made to work, this test fails and becomes the success case above. + */ + @Test + public void dispatchInsideConstructorIsRejectedForLua() { + test().testLua(true).executeProg().expectError("could not be resolved for the Lua target") + .lines(DISPATCH_IN_CONSTRUCTOR); + } + /** Each type argument picks its own instance, so one generic serves several types. */ @Test public void twoInstancesOfOneClass() { From 16f04b83b5a2b455e012e9912b823746d0390c49 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 06:38:33 +0200 Subject: [PATCH 11/21] Say which target a failing Lua test ran on translateAndTestLua never set the environment label, so a Lua failure was reported under whatever Jass configuration ran last: "With Inlining, Optimizations and Stacktraces: Succeed function not called" for a program that had already passed every Jass configuration. This run went looking for a Jass regression that did not exist. Also pins the div/mod result type from both sides rather than changing it. Returning the left operand's type means an integer literal stays assignable to a real through div and mod, where addition collapses two literals to int and does not. That asymmetry is reachable, but OptimizerTests.realFormatting_consistent_ fromIntOps opens with `real a = 1 div 2`, so the suite already defines the current behaviour as correct and tightening it would reject existing code. The backlog carries the question; the tests make sure the answer is deliberate. --- BACKLOG.md | 26 ++++++++++--- .../wurstscript/tests/ExpressionTests.java | 39 +++++++++++++++++++ .../wurstscript/tests/WurstScriptTest.java | 2 + 3 files changed, 62 insertions(+), 5 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 04e4e650c..70157ca84 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -126,6 +126,24 @@ because `LOOP.md` refers to items by number. ## Blocked on a decision +- **8. Should `div` and `mod` keep returning the left operand's type?** Tried returning + `WurstTypeInt.instance()` to match `caseMathOperation` and reverted it: it is a user-visible + breaking change, and the suite already defines the current behaviour as correct. + + The asymmetry is real and reachable. `WurstTypeIntLiteral` is a proper subtype of both int and + real, and `caseMathOperation` collapses two literals to int precisely so `real r = 1 + 1` is an + error. `div`/`mod` return `leftType`, so `real r = 7 div 2` compiles. Changing that made exactly + one test fail — `OptimizerTests.realFormatting_consistent_fromIntOps`, which opens with + `real a = 1 div 2` — and AGENTS.md says the existing suite is the authoritative definition of + behaviour. Real maps will contain the same shape. + + So the question is the owner's: is `real r = 7 div 2` meant to compile? If yes, the branch in + `AttrExprType` wants a comment saying so, and this item closes. If no, it is a deliberate + breaking change that needs the changelog, and `realFormatting_consistent_fromIntOps` needs + rewriting to say what it actually tests, which is real formatting rather than that assignment. + `ExpressionTests.integerDivisionOfLiteralsIsStillAssignableToReal` pins the behaviour meanwhile, + so whichever way it goes is deliberate rather than accidental. + - **Eliminating the remaining `castTo int`.** The motivating case is timer data attachment (`ClosureTimers.wurst`), and the containers behind it: `Table` has 81 casts, `HashList` 13, `HashSet` 6, `HashMap` 4. None can adopt bounds as things stand, because an instance is @@ -142,11 +160,6 @@ because `LOOP.md` refers to items by number. ## Done -- 8. `div` and `mod` return int rather than the left operand's type, matching `caseMathOperation`. - Reachable, not harmless: an integer literal is a proper subtype of both int and real, and - addition collapses two of them to int precisely so `real r = 1 + 1` stays an error — returning - `leftType` skipped that, so `real r = 7 div 2` was accepted. Three tests in `ExpressionTests`: - both operators rejected against a real, and both still int. - 17. A failing Lua test says so. `translateAndTestLua` now sets the environment label instead of reporting under whatever Jass configuration ran last. - 5 (+ the part of 9 that follows it). A type class bound now dispatches from inside a closure on @@ -204,6 +217,9 @@ because `LOOP.md` refers to items by number. - `%` is real modulo in Wurst; `mod` is integer modulo. `int % 8` types as `real`. - Emitted Lua must be byte-identical for identical input (AGENTS.md §8). It is the only emitted output that can be diffed across runs — see item 11. +- The suite is the specification. Before changing what the type checker accepts, grep the tests for + the shape being rejected — item 8 looked like an oversight until one optimizer test turned out to + depend on it. - A test that hangs looks exactly like a test that is slow. If the suite stops making progress, take a thread dump of the forked worker (`jstack `) before killing it — it names the line. - Method names are not what the frontend called them. `LuaDispatchPreparation` renames a whole diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/ExpressionTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/ExpressionTests.java index 6780129a7..e92b3c230 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/ExpressionTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/ExpressionTests.java @@ -486,6 +486,45 @@ private String makeProg(String booleanExpr) { return prog; } + /** + * An integer literal is a proper subtype of both int and real. Addition collapses two of them + * to int, so {@code real r = 1} is allowed while {@code real r = 1 + 1} is not; {@code div} and + * {@code mod} return the left operand's type instead, so a literal stays assignable to a real + * through them. This pins the asymmetry rather than endorsing it — see backlog item 8. + */ + @Test + public void integerDivisionOfLiteralsIsStillAssignableToReal() { + testAssertOkLines(false, + "package test", + "init", + " real quotient = 7 div 2", + " real remainder = 7 mod 2" + ); + } + + @Test + public void additionOfLiteralsIsNotAssignableToReal() { + testAssertErrorsLines(false, "Cannot assign int to real", + "package test", + "init", + " real sum = 7 + 2" + ); + } + + /** Whatever the declared type, both are integer operations at runtime. */ + @Test + public void integerDivisionAndModuloStayInt() { + testAssertOkLines(true, + "package test", + "native testSuccess()", + "init", + " int d = 7 div 2", + " int m = 7 mod 2", + " if d == 3 and m == 1", + " testSuccess()" + ); + } + public void assertOk(String booleanExpr) { String prog = makeProg(booleanExpr); testAssertOk(UtilsIO.getMethodName(1), true, prog); diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java index da1c2a9e7..a3e2712ca 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java @@ -496,6 +496,8 @@ private void testWithoutInliningAndOptimization(String name, boolean executeProg } private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, WurstModel model, WurstCompilerJassImpl compiler) { + // Otherwise a Lua failure is reported under whatever Jass configuration ran last. + setCurrentTestEnv("Lua"); try { name = name.replaceAll("[^a-zA-Z0-9_]", "_"); From a9f5f8eeb649caba840351e8e5eb4cea6e8a409f Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 06:54:30 +0200 Subject: [PATCH 12/21] Resolve a type parameter's default where the binding is known Reading a slot of a T array that was never written gave a stand-in rather than the default of what T stands for, and the stand-in compares equal only to another stand-in - so Box.first() == 0 was quietly false on the interpreter, while both backends had it right. The stand-in is produced by a static attribute, which cannot see the stack frames that know the type argument. ProgramState can, so the value is resolved where it is produced - at the array read and the member read - rather than at the comparison, which is only where the symptom shows. --- BACKLOG.md | 35 ++++++++----------- .../ILconstUnsafeDefault.java | 4 +++ .../interpreter/EvaluateExpr.java | 7 ++-- .../interpreter/ProgramState.java | 17 +++++++++ .../wurstscript/tests/TypeClassTests.java | 21 +++++++++++ 5 files changed, 61 insertions(+), 23 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index 70157ca84..cb262eb2a 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,26 +13,14 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. -16. **A never-written array of a type parameter reads as nothing, silently.** In the interpreter - only — Jass and Lua both give the type argument's default. `DefaultValue.get(ImTypeVarRef)` - returns `ILconstUnsafeDefault`, whose `isEqualTo` matches only another `ILconstUnsafeDefault`, - so a comparison against the real default is quietly false rather than an error. Repro: - - class Box - private static T array none - static function first() returns T - return none[0] - init - if Box.first() == 0 - testSuccess() - - Passes on every Jass configuration and fails on the pre-transform interpreter run. The plain - `int array` version passes, so this is specific to the type parameter. The interpreter knows - the current type argument (`ProgramState.resolveType`), but `DefaultValue` is a static - attribute with no access to it, and the array's default supplier is bound when the array is - allocated rather than when it is read. Either resolve at read time where the state is in hand, - or make the placeholder throw when used — what it must not do is compare unequal in silence. - Found by `FastHashMapTests`: the tombstone fixture needs a "no value" for `V`. +18. **`ILconstUnsafeDefault` can still escape silently.** Item 16 resolves it where a value is + read out of an array or a member, which is how it reached a program. It is produced by + `DefaultValue.get(ImTypeVarRef)` and could surface elsewhere — a return value, a local that was + never assigned — and `isEqualTo` still answers "equal" only for another stand-in, so any path + that is missed stays quiet. Two ways to close it: resolve in `ILInterpreter` where a return + value is defaulted as well, or make `isEqualTo` throw when compared against a real value, so a + missed path fails loudly instead. The second is the stronger check and the riskier change; try + it and see what the suite says. 15. **One junk dispatch slot per specialised class.** Left over from item 3, same heuristic in the other place it is used. `addDirectAliases` composes `owner.getName() + "_" + @@ -160,6 +148,13 @@ because `LOOP.md` refers to items by number. ## Done +- 16. A never-written slot of a `T array` reads as the default of what T stands for. The default + is computed by a static attribute, which cannot see the frames that know the type argument, so + it produced a stand-in that compares equal only to another stand-in — `Box.first() == 0` + was quietly false on the interpreter while both backends had it right. `ProgramState` does know + the substitution, so the stand-in is now resolved where the value is produced, at the array read + and the member read, rather than at the comparison where the symptom shows. Item 18 covers the + paths that could still leak one. - 17. A failing Lua test says so. `translateAndTestLua` now sets the environment label instead of reporting under whatever Jass configuration ran last. - 5 (+ the part of 9 that follows it). A type class bound now dispatches from inside a closure on diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java index ca1a28f39..df9012aaa 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java @@ -18,6 +18,10 @@ public String print() { return "unsafe-default<" + typeVariable.getName() + ">"; } + public ImTypeVar getTypeVariable() { + return typeVariable; + } + public WurstType getType() { return WurstTypeInfer.instance(); } diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/EvaluateExpr.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/EvaluateExpr.java index b4ef93be9..df15e6d64 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/EvaluateExpr.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/EvaluateExpr.java @@ -223,9 +223,9 @@ public static ILconst eval(ImVarArrayAccess e, ProgramState globalState, LocalSt } if (e.getVar().isGlobal()) { - return notNull(globalState.getArrayVal(e.getVar(), indexes), e.getVar().getType(), "Variable " + e.getVar().getName() + " is null.", false); + return globalState.resolveDefault(notNull(globalState.getArrayVal(e.getVar(), indexes), e.getVar().getType(), "Variable " + e.getVar().getName() + " is null.", false)); } else { - return notNull(localState.getArrayVal(e.getVar(), indexes), e.getVar().getType(), "Variable " + e.getVar().getName() + " is null.", false); + return globalState.resolveDefault(notNull(localState.getArrayVal(e.getVar(), indexes), e.getVar().getType(), "Variable " + e.getVar().getName() + " is null.", false)); } } @@ -292,7 +292,8 @@ public static ILconst eval(ImMemberAccess ma, ProgramState globalState, LocalSta Integer val = ((ILconstInt) i.evaluate(globalState, localState)).getVal(); indexes.add(val); } - return receiver.get(ma.getVar(), indexes).orElseGet(() -> ma.attrTyp().defaultValue()); + return globalState.resolveDefault( + receiver.get(ma.getVar(), indexes).orElseGet(() -> ma.attrTyp().defaultValue())); } public static ILconst eval(ImAlloc e, ProgramState globalState, LocalState localState) { diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/ProgramState.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/ProgramState.java index d89f0b376..9a36bb7b7 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/ProgramState.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/interpreter/ProgramState.java @@ -395,6 +395,23 @@ public ImType resolveType(ImType t) { return resolveTypeDeep(t, 32); // small budget to avoid cycles } + /** + * Replaces the stand-in default of a type parameter with the default of the type bound to it. + *

+ * The default of a value is computed by a static attribute, which cannot see the frames that + * know what the parameter stands for, so it produces a stand-in. Reading a slot of a + * {@code T array} that was never written is how one reaches a program: the stand-in compares + * equal only to another stand-in, so a comparison against the real default is quietly false. + * The frames are known here, so resolve it where the value is produced. + */ + public ILconst resolveDefault(ILconst value) { + if (!(value instanceof ILconstUnsafeDefault unsafeDefault)) { + return value; + } + ImType resolved = resolveType(JassIm.ImTypeVarRef(unsafeDefault.getTypeVariable())); + return resolved instanceof ImTypeVarRef ? value : resolved.defaultValue(); + } + private ImType resolveTypeDeep(ImType t, int budget) { if (budget <= 0 || t == null) return t; diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index d077d2707..656aabfa7 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -161,6 +161,27 @@ public void dispatchInsideConstructorIsRejectedForLua() { .lines(DISPATCH_IN_CONSTRUCTOR); } + /** + * A slot of a {@code T array} that was never written reads as the default of whatever T stands + * for. The default is computed by a static attribute, which cannot see what T is bound to, so + * it produces a stand-in — and a stand-in compares equal only to another stand-in, which made + * this quietly false on the interpreter while both backends had it right. + */ + @Test + public void unwrittenArrayOfATypeParameterReadsAsItsDefault() { + testAssertOkLines(true, + "package test", + "native testSuccess()", + "class Box", + " private static T array none", + " static function first() returns T", + " return none[0]", + "init", + " if Box.first() == 0 and Box.first() == null", + " testSuccess()" + ); + } + /** Each type argument picks its own instance, so one generic serves several types. */ @Test public void twoInstancesOfOneClass() { From 1df9e6ef4bddd02bf084f4d52db42486d9d94266 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 07:08:28 +0200 Subject: [PATCH 13/21] Refuse to compare an unresolved type parameter default The stand-in for the default of a type parameter answered "equal" only for another stand-in, so a comparison against a concrete value was quietly false. The previous commit resolves it where a value is read, which is the path that reached a program, but the stand-in is produced by a static attribute and could surface anywhere - and silence was the part worth removing. It now throws when compared to a real value, naming the type parameter. The whole suite passes with it, so nothing reachable produces one any more; if something starts to, it says so rather than returning a wrong answer. --- BACKLOG.md | 14 +++++--------- .../intermediatelang/ILconstUnsafeDefault.java | 11 ++++++++++- 2 files changed, 15 insertions(+), 10 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index cb262eb2a..e82a3f38c 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,15 +13,6 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. -18. **`ILconstUnsafeDefault` can still escape silently.** Item 16 resolves it where a value is - read out of an array or a member, which is how it reached a program. It is produced by - `DefaultValue.get(ImTypeVarRef)` and could surface elsewhere — a return value, a local that was - never assigned — and `isEqualTo` still answers "equal" only for another stand-in, so any path - that is missed stays quiet. Two ways to close it: resolve in `ILInterpreter` where a return - value is defaulted as well, or make `isEqualTo` throw when compared against a real value, so a - missed path fails loudly instead. The second is the stronger check and the riskier change; try - it and see what the suite says. - 15. **One junk dispatch slot per specialised class.** Left over from item 3, same heuristic in the other place it is used. `addDirectAliases` composes `owner.getName() + "_" + semanticNameFromMethodName(name)`, and for a specialised method that trailing segment is the @@ -148,6 +139,11 @@ because `LOOP.md` refers to items by number. ## Done +- 18. Comparing an unresolved type parameter default is now an error rather than a quiet "not + equal". Item 16 closed the path that reached a program, but the stand-in is produced by a static + attribute and could surface anywhere, so the silence was the part worth removing. The whole suite + is green with it throwing, which says nothing reachable produces one any more — and if something + starts to, it says so instead of returning a wrong answer. - 16. A never-written slot of a `T array` reads as the default of what T stands for. The default is computed by a static attribute, which cannot see the frames that know the type argument, so it produced a stand-in that compares equal only to another stand-in — `Box.first() == 0` diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java index df9012aaa..f4a0392a1 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java @@ -1,5 +1,6 @@ package de.peeeq.wurstscript.intermediatelang; +import de.peeeq.wurstio.jassinterpreter.InterpreterException; import de.peeeq.wurstscript.jassIm.ImTypeVar; import de.peeeq.wurstscript.types.WurstType; import de.peeeq.wurstscript.types.WurstTypeInfer; @@ -28,7 +29,15 @@ public WurstType getType() { @Override public boolean isEqualTo(ILconst other) { - return other instanceof ILconstUnsafeDefault; + if (other instanceof ILconstUnsafeDefault) { + return true; + } + // Answering "not equal" is how this stays quiet. The value is the default of a type + // parameter whose binding was not known where the value was produced, so it is not + // comparable to anything concrete. ProgramState.resolveDefault replaces it wherever the + // frames that know the binding are in reach; getting here means a path it does not cover. + throw new InterpreterException("The default value of type parameter " + typeVariable.getName() + + " is not known here, so it cannot be compared to " + other.print() + "."); } } From 9d6b00579a24389b97caa80c77e99c92f0db4acd Mon Sep 17 00:00:00 2001 From: Frotty Date: Sat, 15 Aug 2026 07:28:36 +0200 Subject: [PATCH 14/21] Record why the junk dispatch slot stays for now Tried building the class-qualified dispatch alias from the declared name instead of the trailing segment of the mangled one, to stop every method of a specialised class claiming the same slot. The suite rejected it: overloads share a declared name, so setup(int) and setup(string) collapse into one slot, which is what two LuaTranslationTests cases exist to prevent. Both sources are wrong in opposite directions - the mangled segment collides across the siblings of one specialisation, the declared name across overloads. The junk slot is unused, so leaving it costs one table key per specialised class, while getting the alias wrong costs a real mis-binding. The backlog says what a fix would have to separate. --- BACKLOG.md | 49 +++++++++++++++++++++++++++++++++---------------- 1 file changed, 33 insertions(+), 16 deletions(-) diff --git a/BACKLOG.md b/BACKLOG.md index e82a3f38c..420406814 100644 --- a/BACKLOG.md +++ b/BACKLOG.md @@ -13,14 +13,23 @@ Notes rather than leaving it in a commit message. Numbering is stable: finished items leave a gap rather than shifting the ones below, because `LOOP.md` refers to items by number. -15. **One junk dispatch slot per specialised class.** Left over from item 3, same heuristic in - the other place it is used. `addDirectAliases` composes `owner.getName() + "_" + +15. **One junk dispatch slot per specialised class.** `addDirectAliases` and + `LuaTranslator.collectDispatchSlotNames` both compose `owner.getName() + "_" + semanticNameFromMethodName(name)`, and for a specialised method that trailing segment is the - type argument, so every method of `FastHashMap` claims the same - `FastHashMap_specialized_integer__integer_integer` slot and the alphabetically first wins. - Nothing calls it, so it is dead weight rather than a wrong result — but it is the same - mistake, and the alias it *should* produce is the class qualified with the declared name. - Fixing it changes emitted slot names, so it wants its own commit and its own suite run. + type argument — so every method of `FastHashMap` claims one shared + `FastHashMap_specialized_integer__integer_integer` slot and the alphabetically first wins it. + Nothing calls it, so it is dead weight rather than a wrong result. + + Tried using the declared name instead and reverted it: overloads share a declared name, so + `setup(int)` and `setup(string)` collapse into one slot, which is what + `LuaTranslationTests.overloadedMethodsDoNotAliasInLuaDispatchTables` and + `moduleProvidedOverloadedOverrideDoesNotCollapseLuaSlots` exist to prevent. Both sources of a + semantic name are wrong, in opposite directions: the mangled trailing segment collides across + the siblings of one specialisation, the declared name collides across overloads. A fix needs a + name that separates both — the declared name together with the dispatch signature key would, + since that is already what distinguishes overloads elsewhere in the same file. Worth doing only + if this stops being dead weight, because the cost of getting it wrong is a real mis-binding + while the cost of leaving it is one unused table key per specialised class. 6. **Lua dispatch inside the constructor** of a bounded generic class. Works on Jass; there is now a repro for both targets, `TypeClassTests.dispatchInsideConstructor` and @@ -34,15 +43,20 @@ because `LOOP.md` refers to items by number. outermost one a concrete argument. `collectGenericNewUse` requires non-empty type arguments, so it never starts. - The instantiation is only on the type of what the call is assigned to. Three ways to get at it, - roughly in order of how much they would disturb: attach the class's type arguments to - constructor calls when the intermediate language is built, which is where the frontend still - knows them and would serve both targets uniformly — but it changes the Jass path, which reaches - the same answer another way today, so the emitted `.j` needs checking; read them from the - assignment target on the Lua path, which is a syntactic shape and would miss - `foo(new Box(21))`; or specialise from the `#alloc` inside the constructor, which is the - item 5 mechanism but would have to reach back out to the caller. The first looks right; confirm - it is what the Jass path already relies on before changing it. + What Jass does, from `TypeClassTests_dispatchInsideConstructor_no_opts.jim`: it specialises the + constructor function itself, `b_8 = new_Box⟪integer⟫(21)`. It gets there from *types*, not from + the call — `collectGenericUsages` collects a `GenericVar` for the local declared + `Box` and a `GenericReturnTypeFunc` for `new_Box`, whose return type is generic. + The Lua collector has neither; it only ever looks at calls. So attaching type arguments to + constructor calls, which an earlier note here proposed, is not what the Jass path relies on and + would be a second mechanism rather than the same one. + + The honest next step is to collect from types on the Lua path too, restricted the way item 5's + collection is. That runs straight into the same design question, though: `GenericVar` and + `GenericReturnTypeFunc` specialise the *class*, and item 5 showed that an object coming from a + specialised class while its methods are bound to the erased one breaks everything. Either the + collection has to specialise only the constructor path and leave the object erased, or Lua stops + erasing constructed generic classes — which is a decision about the erasure model, not a patch. 7. **Module bounds.** `module M` is rejected with a clear message today. Needs receiver rewriting during expansion, or type parameters on `ModuleInstanciation`. @@ -208,6 +222,9 @@ because `LOOP.md` refers to items by number. - `%` is real modulo in Wurst; `mod` is integer modulo. `int % 8` types as `real`. - Emitted Lua must be byte-identical for identical input (AGENTS.md §8). It is the only emitted output that can be diffed across runs — see item 11. +- Two of this run's reverts were the same mistake: a name that looks redundant is usually carrying + a distinction. The mangled method name separates overloads; `leftType` on `div` keeps a literal + assignable to a real. Check what a name distinguishes before replacing it with a tidier one. - The suite is the specification. Before changing what the type checker accepts, grep the tests for the shape being rejected — item 8 looked like an oversight until one optimizer test turned out to depend on it. From 7871731728d95877eb41daef0e15654ead7c5e53 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 08:37:19 +0200 Subject: [PATCH 15/21] Reject Lua keywords when validating a generated name Wurst and Lua reserve different words, so a method can be declared repeat or goto and reach the backend under that name. Method names survive it, because the pass that assigns them uniques against the reserved set. A closure does not: it adds the name it implements as a dispatch alias directly, so the alias arrives as a bare keyword and is emitted as a table key. luac rejects that, so it was loud rather than wrong, but the check added alongside it accepted the name - and catching this before the syntax check is the whole point of having it. isValid now rejects keywords and toIdentifier maps them out of the way. Underscores rather than a counter, so a keyword maps to the same name wherever it is derived: call sites and class tables have to agree without consulting each other. --- .../lua/translation/LuaIdentifiers.java | 17 ++++++++++++++-- .../wurstscript/tests/TypeClassTests.java | 20 +++++++++++++++++++ 2 files changed, 35 insertions(+), 2 deletions(-) diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java index 95cde0a8b..b0da4381f 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/lua/translation/LuaIdentifiers.java @@ -13,7 +13,14 @@ public final class LuaIdentifiers { /** Lua's vararg parameter, which is a legal parameter name but not an identifier. */ public static final String VARARG = "..."; - /** Whether {@code name} can be used as-is as a Lua identifier or table key. */ + /** + * Whether {@code name} can be used as-is as a Lua identifier or table key. + * + *

A keyword is spelled like an identifier and is not one. Wurst reserves a different set, so + * a method can be declared {@code repeat} or {@code goto} and reach the backend under that + * name; emitted as a table key it is a syntax error rather than a wrong result, but this is the + * check that is supposed to catch it first. + */ public static boolean isValid(String name) { if (name == null || name.isEmpty() || isDigit(name.charAt(0))) { return false; @@ -23,7 +30,7 @@ public static boolean isValid(String name) { return false; } } - return true; + return !LuaReservedNames.LUA_KEYWORDS.contains(name); } /** Maps any name onto a Lua identifier, leaving names that already are one untouched. */ @@ -39,6 +46,12 @@ public static String toIdentifier(String name) { if (sb.length() == 0 || isDigit(sb.charAt(0))) { sb.insert(0, '_'); } + // A trailing underscore rather than a counter, so the name a keyword maps to is the same + // wherever it is derived - call sites and class tables have to agree without consulting + // each other. + while (LuaReservedNames.LUA_KEYWORDS.contains(sb.toString())) { + sb.append('_'); + } return sb.toString(); } diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index 6e0ba186c..cd8429a4e 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -121,6 +121,26 @@ public void dispatchInsideClosureIsRejectedForLua() { ); } + /** + * Wurst and Lua reserve different words, so a method can be declared {@code repeat} and reach + * the backend under that name. A closure adds the name it implements as a dispatch alias + * directly, without the uniquing that protects method names, so the alias arrives as a bare + * keyword and is emitted as a table key — {@code expected near 'repeat'} from luac. + */ + @Test + public void closureImplementingALuaKeywordName() { + test().testLua(true).executeProg().lines( + "package test", + "native testSuccess()", + "interface Producer", + " function repeat() returns int", + "init", + " Producer p = () -> 42", + " if p.repeat() == 42", + " testSuccess()" + ); + } + /** Each type argument picks its own instance, so one generic serves several types. */ @Test public void twoInstancesOfOneClass() { From c1ac493a7bb17634e6c54617007dfbf6afc07532 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 08:45:51 +0200 Subject: [PATCH 16/21] Cover the bounded Lua runner, and cap what it keeps Two tests for the runner itself, both Lua only: run through the Jass configurations the non-terminating one would hang the interpreter instead, which is the same problem somewhere this fix does not reach. - a program that loops forever fails its own test rather than the run - a program that prints twenty thousand lines still finishes The second is the deadlock this change was for: with the streams read one after the other, the program blocks writing stdout while the runner blocks reading stderr. The timeout is overridable so these take seconds rather than a minute. Output kept for the failure message is capped. Draining still never stops, since stopping is what blocks the process, but a program that loops while printing would otherwise exhaust the worker before the timeout fires. Capping meant success could no longer be read back out of the retained text - testSuccess prints last, well past the limit - so the line is recognised while draining. Which is better regardless: whether a program succeeded no longer depends on how much of its output was kept. --- .../wurstscript/tests/LuaRunnerTests.java | 63 +++++++++++++++++++ .../wurstscript/tests/WurstScriptTest.java | 54 ++++++++++++---- 2 files changed, 106 insertions(+), 11 deletions(-) create mode 100644 de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/LuaRunnerTests.java diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/LuaRunnerTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/LuaRunnerTests.java new file mode 100644 index 000000000..a719745b5 --- /dev/null +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/LuaRunnerTests.java @@ -0,0 +1,63 @@ +package tests.wurstscript.tests; + +import org.testng.annotations.Test; + +import static org.testng.Assert.assertTrue; +import static org.testng.Assert.fail; + +/** + * Tests for the harness that runs emitted Lua, rather than for anything it compiles. + *

+ * Both cases here used to stop the whole suite instead of failing a test: the runner drained the + * spawned interpreter's stderr to the end before reading stdout, and waited for the process without + * a bound. A program that fills the stdout pipe blocks writing while the runner blocks reading the + * other pipe, and neither side ever moves. There is nothing to see when that happens - no output, + * no failing test, no timeout - so it is worth holding onto tests that would notice it coming back. + *

+ * Both are Lua only. Run through the Jass configurations as well, the non-terminating one would + * hang the interpreter instead, which is the same problem in a place this fix does not reach. + */ +public class LuaRunnerTests extends WurstScriptTest { + + /** Long enough that a program which does terminate still does, short enough to wait for. */ + @Override + protected int luaExecutionTimeoutSeconds() { + return 5; + } + + @Test + public void aProgramThatDoesNotTerminateFailsItsOwnTest() { + try { + test().testLua(true).luaOnly(true).executeProg().lines( + "package test", + "native testSuccess()", + "init", + " var i = 0", + " while true", + " i += 1", + " testSuccess()" + ); + fail("expected the runner to give up on a program that does not terminate"); + } catch (Error e) { + assertTrue(e.getMessage() != null && e.getMessage().contains("did not terminate"), + "expected a timeout, got: " + e.getMessage()); + } + } + + /** + * Enough lines to fill the pipe buffer several times over. With the streams read one after the + * other this deadlocks: the program blocks writing stdout, the runner blocks reading stderr. + */ + @Test + public void aProgramThatFloodsStdoutStillFinishes() { + test().testLua(true).luaOnly(true).executeProg().lines( + "package test", + "native testSuccess()", + "native println(string s)", + "init", + " for i = 1 to 20000", + " println(\"filling the pipe buffer with a reasonably long line of output\")", + " testSuccess()" + ); + } +} diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java index da1c2a9e7..320a9a375 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/WurstScriptTest.java @@ -51,6 +51,18 @@ public class WurstScriptTest { /** Generous enough for the slowest test program, short enough that a loop fails the run. */ private static final int LUA_EXECUTION_TIMEOUT_SECONDS = 60; + + /** + * How much of a program's output is kept for the failure message. Draining never stops - that + * is what deadlocks the run - but retaining everything from a program that loops while printing + * would exhaust the worker before the timeout fires. + */ + private static final int RETAINED_OUTPUT_LIMIT = 64 * 1024; + + /** Overridden by the tests covering the runner itself, so they need seconds rather than a minute. */ + protected int luaExecutionTimeoutSeconds() { + return LUA_EXECUTION_TIMEOUT_SECONDS; + } private static volatile String resolvedLuaExecutable; private static volatile String resolvedLuacExecutable; private static volatile String extractedLuaWin; @@ -555,12 +567,14 @@ private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, // Both pipes must be drained while the program runs, and the wait must end: a // generated program that loops forever would otherwise hang the whole suite, // and one that fills the stdout pipe would deadlock against a stderr-first read. - Thread outCollector = collectStreamAsync(p.getInputStream(), output); + java.util.concurrent.atomic.AtomicBoolean sawTestSuccess = + new java.util.concurrent.atomic.AtomicBoolean(false); + Thread outCollector = collectStreamAsync(p.getInputStream(), output, "testSuccess", sawTestSuccess); Thread errCollector = collectStreamAsync(p.getErrorStream(), errors); - if (!p.waitFor(LUA_EXECUTION_TIMEOUT_SECONDS, TimeUnit.SECONDS)) { + if (!p.waitFor(luaExecutionTimeoutSeconds(), TimeUnit.SECONDS)) { p.destroyForcibly(); throw new Error(currentTestEnv + ": Lua program did not terminate within " - + LUA_EXECUTION_TIMEOUT_SECONDS + "s: " + luaFile.getName()); + + luaExecutionTimeoutSeconds() + "s: " + luaFile.getName()); } outCollector.join(); errCollector.join(); @@ -570,13 +584,7 @@ private void translateAndTestLua(String name, boolean executeProg, WurstGui gui, throw new TestFailException(errors.toString()); } - boolean success = false; - for (String outputLine : output.toString().split("\n", -1)) { - if (outputLine.equals("testSuccess")) { - success = true; - } - } - if (!success) { + if (!sawTestSuccess.get()) { throw new Error(currentTestEnv + ": Succeed function not called"); } } @@ -625,11 +633,35 @@ private void checkLuaSyntax(String luacExecutable, File luaFile) throws IOExcept } private Thread collectStreamAsync(InputStream stream, StringBuilder out) { + return collectStreamAsync(stream, out, null, null); + } + + /** + * Drains {@code stream} into {@code out}, keeping only the first {@link #RETAINED_OUTPUT_LIMIT} + * characters. Reading never stops — that is what blocks the process and deadlocks the run — but + * a program that loops while printing would fill the heap before the timeout could fire. + *

+ * Whatever the caller is looking for is recognised here rather than read back out of + * {@code out} afterwards, because it may arrive after the limit: a program that prints a + * million lines and then succeeds has still succeeded. + */ + private Thread collectStreamAsync(InputStream stream, StringBuilder out, + String watchedLine, java.util.concurrent.atomic.AtomicBoolean sawWatchedLine) { Thread t = new Thread(() -> { try (BufferedReader input = new BufferedReader(new InputStreamReader(stream))) { String line; + boolean truncated = false; while ((line = input.readLine()) != null) { - out.append(line).append("\n"); + if (watchedLine != null && watchedLine.equals(line)) { + sawWatchedLine.set(true); + } + if (out.length() < RETAINED_OUTPUT_LIMIT) { + out.append(line).append("\n"); + } else if (!truncated) { + truncated = true; + out.append("... further output dropped after ") + .append(RETAINED_OUTPUT_LIMIT).append(" characters\n"); + } } } catch (IOException ignored) { } From 9f16d43c8d364cef01d32cf0fa018bb1cfc7dfc4 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 08:47:58 +0200 Subject: [PATCH 17/21] Check the full-table sentinel in put as well slotFor says a key has nowhere to go by returning a slot below the instance's section, and get, has and remove all check for it. put did not, so a ninth distinct key wrote through the sentinel: an entry nothing can reach, a count of nine, and a negative array index on targets less forgiving than Lua. The test fills the map exactly - keys one to eight hash to the eight slots - and puts a ninth. It does not read the absent key back, because reading one of a type parameter returns a stand-in the interpreter cannot compare; that is a separate bug, fixed later in this stack. --- .../wurstscript/tests/FastHashMapTests.java | 26 +++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java index 4b4938434..23817b5a2 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/FastHashMapTests.java @@ -72,6 +72,10 @@ private static String[] fastHashMap(String... extra) { " return firstDead", " function put(K key, V value)", " let s = slotFor(key)", + // Capacity is fixed, so a new key can arrive with nowhere to go. The other three + // guard the same way; without it this writes through the sentinel and counts it. + " if s < base", + " return", " if not used[s]", " used[s] = true", " dead[s] = false", @@ -201,6 +205,28 @@ private static boolean namesMethod(String name, String method) { " testSuccess()", }; + /** + * Capacity is fixed, so the ninth distinct key has nowhere to go: keys 1 to 8 hash to the eight + * slots exactly. `slotFor` says so by returning a slot below this instance's section, and every + * entry point has to check that before indexing — otherwise the write goes through the sentinel, + * the entry is unreachable, and the count says nine. + */ + @Test + public void puttingIntoAFullMapChangesNothing() { + testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, new String[]{ + "init", + " let m = new FastHashMap()", + " for i = 1 to 8", + " m.put(i, i * 10)", + " m.put(9, 90)", + " if m.size() == 8 and not m.has(9)", + // Not m.get(9) as well: reading an absent key of a type parameter returns a stand-in + // the interpreter cannot compare, which is a separate bug fixed later in this stack. + " if m.get(1) == 10 and m.get(8) == 80", + " testSuccess()" + })); + } + @Test public void removeLeavesATombstone() { testAssertOkLines(true, program(fastHashMap(), INT_INSTANCE, USE_WITH_REMOVE)); From c5d9dfb2d4c3d13fc5a54b8a4c3e874356cfca34 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 08:53:59 +0200 Subject: [PATCH 18/21] Follow calls when deciding a closure needs its construction read The gate asked whether a dispatch sits in the closure itself. A closure whose body is () -> helper(x) has none - the dispatch is in helper - so the construction was skipped, and the bound reached the backend unresolved even though this is exactly the shape the change was for. It now uses the call-aware check that already exists for functions. classNeedsSpecialization is left as it was, because widening that would change what gets specialised on paths which have nothing to do with closures. Specialising the closure class then surfaced a second thing: the member access collector ran again for the copy, and a receiver still typed by the generic class cannot adapt to the specialised one, which is not a superclass of it. A class with no type variables left has nothing to select, so it is skipped. --- .../imtranslation/EliminateGenerics.java | 33 +++++++++++++++++-- .../wurstscript/tests/TypeClassTests.java | 28 ++++++++++++++++ 2 files changed, 59 insertions(+), 2 deletions(-) diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java index 18d0a83fe..43d4dca0b 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/translation/imtranslation/EliminateGenerics.java @@ -278,7 +278,33 @@ private void collectGenericNewUse(ImAlloc alloc) { * to the erased one. */ private boolean isConstructionOnlyInstantiation(ImClass classDef) { - return classDef.attrTrace() instanceof ExprClosure && classNeedsSpecialization(classDef); + return classDef.attrTrace() instanceof ExprClosure && classReachesDispatch(classDef); + } + + /** + * Whether anything the class does ends in a dispatch on a bound, including through the + * functions it calls. `classNeedsSpecialization` asks only whether a dispatch sits in the class + * itself, which is the wrong question here: a closure whose body is `() -> helper(x)` has no + * dispatch of its own, and the instantiation it needs is still only known at its construction. + * That question is kept as it is, because widening it would change what gets specialised on + * paths that have nothing to do with closures. + */ + private boolean classReachesDispatch(ImClass classDef) { + for (ImFunction f : classDef.getFunctions()) { + if (functionNeedsSpecialization(f, Collections.newSetFromMap(new IdentityHashMap<>()), + Collections.newSetFromMap(new IdentityHashMap<>()))) { + return true; + } + } + for (ImMethod m : classDef.getMethods()) { + if (m.getImplementation() != null + && functionNeedsSpecialization(m.getImplementation(), + Collections.newSetFromMap(new IdentityHashMap<>()), + Collections.newSetFromMap(new IdentityHashMap<>()))) { + return true; + } + } + return false; } /** @@ -291,7 +317,10 @@ private void collectGenericNewUse(ImMemberAccess memberAccess) { if (field.getParent() == null || !(field.getParent().getParent() instanceof ImClass owningClass)) { return; } - if (!isConstructionOnlyInstantiation(owningClass)) { + // A class that has already been specialised has nothing left to select, and asking the + // receiver to adapt to it fails outright: the receiver is still typed by the generic class + // the specialised one was copied from, which is not a superclass of it. + if (owningClass.getTypeVariables().isEmpty() || !isConstructionOnlyInstantiation(owningClass)) { return; } if (memberAccess.getTypeArguments().isEmpty()) { diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index d077d2707..907aba6d4 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -161,6 +161,34 @@ public void dispatchInsideConstructorIsRejectedForLua() { .lines(DISPATCH_IN_CONSTRUCTOR); } + /** + * The closure has no dispatch of its own — it calls something that does. The gate deciding + * whether a construction is the only place an instantiation is stated has to follow calls to + * see that, or this reaches the backend with the bound unresolved. + */ + @Test + public void dispatchInsideClosureThroughHelperLua() { + test().testLua(true).executeProg().lines( + "package test", + "native testSuccess()", + "interface ToIndex", + " function toIndex(T x) returns int", + "implements ToIndex", + " function toIndex(int x) returns int", + " return x * 2", + "interface Producer", + " function produce() returns int", + "function helper(Q x) returns int", + " return Q.toIndex(x)", + "function foo(Q x) returns int", + " Producer p = () -> helper(x)", + " return p.produce()", + "init", + " if foo(21) == 42", + " testSuccess()" + ); + } + /** Each type argument picks its own instance, so one generic serves several types. */ @Test public void twoInstancesOfOneClass() { From 7582ade618ba4367fda6105dbb9fe9d4e69b23d2 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 08:58:58 +0200 Subject: [PATCH 19/21] Refuse the unresolved default whichever side it is on Equality asks the left operand whether it equals the right, so putting the refusal in the stand-in only covered half of it: unresolved == 0 complained while 0 == unresolved went quietly false, which is the failure the refusal was added to remove. The check moves to the operator, where both operands are in hand, and the stand-in goes back to answering plainly. Two stand-ins still compare equal to each other; only a comparison against a real value is refused. --- .../de/peeeq/wurstscript/WurstOperator.java | 37 +++++++++++++++++-- .../ILconstUnsafeDefault.java | 14 ++----- .../wurstscript/tests/TypeClassTests.java | 19 ++++++++++ 3 files changed, 55 insertions(+), 15 deletions(-) diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/WurstOperator.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/WurstOperator.java index d9e27d8b3..b298a8702 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/WurstOperator.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/WurstOperator.java @@ -1,6 +1,7 @@ package de.peeeq.wurstscript; import de.peeeq.wurstscript.attributes.AttrFuncDef; +import de.peeeq.wurstio.jassinterpreter.InterpreterException; import de.peeeq.wurstscript.intermediatelang.*; import de.peeeq.wurstscript.jassAst.JassAst; import de.peeeq.wurstscript.jassAst.JassOpBinary; @@ -129,6 +130,28 @@ public LuaOpBinary luaTranslateBinary() { throw new Error("cannot translate " + this); } + /** + * Refuses to compare the stand-in for a type parameter's default, whichever side it is on. + *

+ * The stand-in exists because the default of a value is computed by a static attribute, which + * cannot see what the parameter is bound to. It answers "equal" only for another stand-in, so + * comparing one against a real value is a wrong answer rather than an error. Doing this here + * rather than in the value itself keeps it independent of operand order: only the left operand + * gets asked, so `0 == unresolved` would otherwise go quietly false while `unresolved == 0` + * complained. + */ + private static void rejectUnresolvedDefault(ILconst left, ILconst right) { + ILconstUnsafeDefault unresolved = left instanceof ILconstUnsafeDefault leftDefault ? leftDefault + : right instanceof ILconstUnsafeDefault rightDefault ? rightDefault : null; + if (unresolved == null || (left instanceof ILconstUnsafeDefault && right instanceof ILconstUnsafeDefault)) { + return; + } + throw new InterpreterException("The default value of type parameter " + + unresolved.getTypeVariable().getName() + + " is not known here, so it cannot be compared to " + + (unresolved == left ? right : left).print() + "."); + } + public ILconst evaluateBinaryOperator(ILconst left, Supplier right) { switch (this) { @@ -140,8 +163,11 @@ public ILconst evaluateBinaryOperator(ILconst left, return new ILconstInt(((ILconstInt) left).getVal() / ((ILconstInt) right.get()).getVal()); case DIV_REAL: return new ILconstReal(getReal(left) / getReal(right.get())); - case EQ: - return ILconstBool.instance(left.equals(right.get())); + case EQ: { + ILconst rightVal = right.get(); + rejectUnresolvedDefault(left, rightVal); + return ILconstBool.instance(left.equals(rightVal)); + } case GREATER: return ((ILconstNum) left).greater((ILconstNum) right.get()); case GREATER_EQ: @@ -160,8 +186,11 @@ public ILconst evaluateBinaryOperator(ILconst left, return new ILconstReal(moduloReal(getReal(left), getReal(right.get()))); case MULT: return ((ILconstNum) left).mul((ILconstNum) right.get()); - case NOTEQ: - return ILconstBool.instance(!left.equals(right.get())); + case NOTEQ: { + ILconst rightVal = right.get(); + rejectUnresolvedDefault(left, rightVal); + return ILconstBool.instance(!left.equals(rightVal)); + } case PLUS: return ((ILconstAddable) left).add((ILconstAddable) right.get()); case NOT: diff --git a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java index f4a0392a1..52cfbbb53 100644 --- a/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java +++ b/de.peeeq.wurstscript/src/main/java/de/peeeq/wurstscript/intermediatelang/ILconstUnsafeDefault.java @@ -1,6 +1,5 @@ package de.peeeq.wurstscript.intermediatelang; -import de.peeeq.wurstio.jassinterpreter.InterpreterException; import de.peeeq.wurstscript.jassIm.ImTypeVar; import de.peeeq.wurstscript.types.WurstType; import de.peeeq.wurstscript.types.WurstTypeInfer; @@ -29,15 +28,8 @@ public WurstType getType() { @Override public boolean isEqualTo(ILconst other) { - if (other instanceof ILconstUnsafeDefault) { - return true; - } - // Answering "not equal" is how this stays quiet. The value is the default of a type - // parameter whose binding was not known where the value was produced, so it is not - // comparable to anything concrete. ProgramState.resolveDefault replaces it wherever the - // frames that know the binding are in reach; getting here means a path it does not cover. - throw new InterpreterException("The default value of type parameter " + typeVariable.getName() - + " is not known here, so it cannot be compared to " + other.print() + "."); + // Comparing this against a real value is refused by WurstOperator, which can see both + // operands; doing it here would depend on which side the stand-in happened to land on. + return other instanceof ILconstUnsafeDefault; } - } diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index 97c624c76..87c8f7156 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -210,6 +210,25 @@ public void unwrittenArrayOfATypeParameterReadsAsItsDefault() { ); } + /** + * The same comparison the other way round. Only the left operand is asked whether it is equal, + * so a stand-in on the right would have gone quietly false while one on the left complained. + */ + @Test + public void unwrittenArrayOfATypeParameterReadsAsItsDefaultReversed() { + testAssertOkLines(true, + "package test", + "native testSuccess()", + "class Box", + " private static T array none", + " static function first() returns T", + " return none[0]", + "init", + " if 0 == Box.first() and null == Box.first()", + " testSuccess()" + ); + } + /** Each type argument picks its own instance, so one generic serves several types. */ @Test public void twoInstancesOfOneClass() { From a587e86c960dafc8054eb7631ced79911dd353be Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 09:05:33 +0200 Subject: [PATCH 20/21] Keep local session files out of the repository Neither of these belongs in a branch that gets merged. The loop configuration is machine-local state - an active session id, an iteration count, a path and a branch name from one checkout - and another checkout picking it up would inherit an automation session that has nothing to do with it. The brief beside it is one run's instructions, not documentation of anything. Both came in with the branch this work was based on rather than with the work. Ignored as well, so they do not come back the next time either is written. --- .claude/ralph-loop.local.md | 10 -------- .gitignore | 2 ++ LOOP.md | 48 ------------------------------------- 3 files changed, 2 insertions(+), 58 deletions(-) delete mode 100644 .claude/ralph-loop.local.md delete mode 100644 LOOP.md diff --git a/.claude/ralph-loop.local.md b/.claude/ralph-loop.local.md deleted file mode 100644 index 305f01377..000000000 --- a/.claude/ralph-loop.local.md +++ /dev/null @@ -1,10 +0,0 @@ ---- -active: true -iteration: 1 -session_id: 0413fc86-e9a2-4954-bc47-1620ed81d95e -max_iterations: 0 -completion_promise: null -started_at: "2026-08-15T01:51:05Z" ---- - -Work on the WurstScript compiler at c:Usersrun-eDocumentsGitHubWurstScript, on the branch feat/fasthashmap-proof. Read AGENTS.md and BACKLOG.md before doing anything. diff --git a/.gitignore b/.gitignore index 7b7bf5ad3..fad3ad29c 100644 --- a/.gitignore +++ b/.gitignore @@ -60,3 +60,5 @@ de.peeeq.wurstscript/output.txt /HelperScripts/.gradle /.gradle-user-home /gradle-home-temp +/.claude/ +/LOOP.md diff --git a/LOOP.md b/LOOP.md deleted file mode 100644 index ed3234d69..000000000 --- a/LOOP.md +++ /dev/null @@ -1,48 +0,0 @@ -# Overnight loop brief - -Work on the WurstScript compiler in this repository, on the branch -`feat/fasthashmap-proof`. Read `AGENTS.md` and `BACKLOG.md` before doing anything. - -## Each iteration - -1. Read `BACKLOG.md`. Take the top item that isn't blocked. -2. Do it. Root-cause it — adjust the underlying system rather than patching a symptom. - If it turns out to be bigger than one iteration, split it in the backlog and do the - first part. -3. Verify. Targeted tests while iterating; the full suite before any commit that touches - main source: - - cd de.peeeq.wurstscript && ./gradlew test - - It must be green. Emitted Lua must stay byte-identical unless the change is meant to - alter it — compare two runs to check. Do not diff `.j` across runs; it is not stable, - and backlog item 11 explains why. -4. Commit and push. Small commits, one concern each. Never end an iteration with - uncommitted work. -5. Update `BACKLOG.md`: move finished items to Done with one line on what actually - happened, add anything learned to Notes, reorder if something more urgent turned up. - -## Rules - -- Never force-push, rewrite history, merge, push to master, or open a PR. The branch gets - reviewed in the morning. -- Commits are authored as the repository owner. No AI, assistant, or co-author references - anywhere in commit messages or code comments. -- Never stop to ask. If something needs a decision from the owner, write the question into - `BACKLOG.md` under that item, mark it blocked, and move to the next item. -- If the full suite goes red and it can't be fixed within the iteration, revert the change, - note why in `BACKLOG.md`, and move on. Leave the branch green. -- Comments explain why, not what. Match the surrounding style. -- Do not touch `de.peeeq.wurstscript/temp/WurstStdlib2` — a fetched test artefact, editing - it changes nothing real. - -## Keeping going - -- Finishing an item is not the end of the run. Go back to step 1 and take the next one. Do - not stop to summarise, and do not treat a green suite as a finish line. -- At most one item may be marked blocked per iteration, and only if it genuinely needs a - decision rather than more work. If everything left looks blocked, that is wrong about at - least one of them — re-read and start the one that can be moved furthest. -- If an item goes three iterations without landing a commit, split it in the backlog and - move on. Don't spend the whole run on one thing. -- Backlog item 12 is a standing item that never completes. There is always a next item. From 46ad1a44baa4a3934027d20285b59624bad50253 Mon Sep 17 00:00:00 2001 From: Frotty Date: Sun, 16 Aug 2026 09:11:21 +0200 Subject: [PATCH 21/21] Merge branch 'lua/closure-type-class-dispatch' into interpreter/type-parameter-defaults --- .../wurstscript/tests/TypeClassTests.java | 68 +++++-------------- 1 file changed, 16 insertions(+), 52 deletions(-) diff --git a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java index 53875adac..85a941e4f 100644 --- a/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java +++ b/de.peeeq.wurstscript/src/test/java/tests/wurstscript/tests/TypeClassTests.java @@ -190,32 +190,33 @@ public void dispatchInsideClosureThroughHelperLua() { } /** - * A slot of a {@code T array} that was never written reads as the default of whatever T stands - * for. The default is computed by a static attribute, which cannot see what T is bound to, so - * it produces a stand-in — and a stand-in compares equal only to another stand-in, which made - * this quietly false on the interpreter while both backends had it right. + * Wurst and Lua reserve different words, so a method can be declared {@code repeat} and reach + * the backend under that name. A closure adds the name it implements as a dispatch alias + * directly, without the uniquing that protects method names, so the alias arrives as a bare + * keyword and is emitted as a table key — {@code expected near 'repeat'} from luac. */ @Test - public void unwrittenArrayOfATypeParameterReadsAsItsDefault() { - testAssertOkLines(true, + public void closureImplementingALuaKeywordName() { + test().testLua(true).executeProg().lines( "package test", "native testSuccess()", - "class Box", - " private static T array none", - " static function first() returns T", - " return none[0]", + "interface Producer", + " function repeat() returns int", "init", - " if Box.first() == 0 and Box.first() == null", + " Producer p = () -> 42", + " if p.repeat() == 42", " testSuccess()" ); } /** - * The same comparison the other way round. Only the left operand is asked whether it is equal, - * so a stand-in on the right would have gone quietly false while one on the left complained. + * A slot of a {@code T array} that was never written reads as the default of whatever T stands + * for. The default is computed by a static attribute, which cannot see what T is bound to, so + * it produces a stand-in — and a stand-in compares equal only to another stand-in, which made + * this quietly false on the interpreter while both backends had it right. */ @Test - public void unwrittenArrayOfATypeParameterReadsAsItsDefaultReversed() { + public void unwrittenArrayOfATypeParameterReadsAsItsDefault() { testAssertOkLines(true, "package test", "native testSuccess()", @@ -224,27 +225,7 @@ public void unwrittenArrayOfATypeParameterReadsAsItsDefaultReversed() { " static function first() returns T", " return none[0]", "init", - " if 0 == Box.first() and null == Box.first()", - " testSuccess()" - ); - } - - /** - * Wurst and Lua reserve different words, so a method can be declared {@code repeat} and reach - * the backend under that name. A closure adds the name it implements as a dispatch alias - * directly, without the uniquing that protects method names, so the alias arrives as a bare - * keyword and is emitted as a table key — {@code expected near 'repeat'} from luac. - */ - @Test - public void closureImplementingALuaKeywordName() { - test().testLua(true).executeProg().lines( - "package test", - "native testSuccess()", - "interface Producer", - " function repeat() returns int", - "init", - " Producer p = () -> 42", - " if p.repeat() == 42", + " if Box.first() == 0 and Box.first() == null", " testSuccess()" ); } @@ -264,23 +245,6 @@ public void unwrittenArrayOfATypeParameterReadsAsItsDefaultReversed() { " return none[0]", "init", " if 0 == Box.first() and null == Box.first()", -======= - * Wurst and Lua reserve different words, so a method can be declared {@code repeat} and reach - * the backend under that name. A closure adds the name it implements as a dispatch alias - * directly, without the uniquing that protects method names, so the alias arrives as a bare - * keyword and is emitted as a table key — {@code expected near 'repeat'} from luac. - */ - @Test - public void closureImplementingALuaKeywordName() { - test().testLua(true).executeProg().lines( - "package test", - "native testSuccess()", - "interface Producer", - " function repeat() returns int", - "init", - " Producer p = () -> 42", - " if p.repeat() == 42", ->>>>>>> lua/closure-type-class-dispatch " testSuccess()" ); }