Add explicit integer-width casts and improve pointer arithmetic#125
Merged
Conversation
Add explicit width casts at sites where FPC's "Converting the operands to Int64 before doing the add/subtract/multiply" and "Mixing signed expressions and longwords gives a 64bit result" hints fire because a narrow (32-bit) sub-operation feeds a wider expression. The casts make the intended width explicit and silence the hints on both 32-bit and 64-bit; the computed values are unchanged on every word size. - ClpMlDsaCore.pas (NTT): Int64(LZeta) * (Int64(LA0) - Int64(LA1)) so the coefficient subtract widens before the multiply. - ClpMlKemCore.pas (compress): (Int64(LC shl n) + (MlKemQ shr 8)) so the inner add runs in Int64; (decompress) cast the leading shift operand to UInt32 so the bit-packing OR no longer mixes a signed Int32 with a longword. - ClpPoly1305.pas: (Int64(Int32(H4 shr 26)) - 1) * 5 so the subtract is Int64. - ClpDateTimeUtilities.pas: Int64(LOffsetHours) * 60 + LOffsetMinutes for the UTC offset (IncMinute already takes Int64). Deliberately left untouched: TMod.Inverse32, TX448Field.Sub, and the ML-DSA eta=2 mod-5 reduction. Their hints are 32-bit-only false positives over intentional fixed-width (UInt32) modular arithmetic; FPC flags any UInt32*UInt32 / UInt32+UInt32 on a 32-bit target regardless of casts, and widening to Int64 would be wrong for Inverse32 (it relies on mod-2^32 wraparound). These are left as the original code rather than suppressed.
FPC flags "Conversion between ordinals and pointers is not portable" at sites
that cast a pointer through NativeUInt. These are false positives (NativeUInt is
pointer-sized on all targets, so the conversions are size-safe on Delphi/FPC and
32/64-bit), but the sites that only need address arithmetic or comparison can
drop the integer round-trip entirely, since {$POINTERMATH ON} is global.
- ClpOcbBlockCipher.pas: compute the block-0 lookahead pointer with pointer
arithmetic, PByte(@ainput[AInOff]) - BLOCK_SIZE, instead of
Pointer(NativeUInt(@ainput[AInOff]) - BLOCK_SIZE). Removes the hint and the
warning; same address.
- ClpAesEngineX86.pas (ProcessBlock / ProcessFourBlocks / ProcessEightBlocks):
do the disjoint-vs-overlap test by comparing the PByte arguments directly
(AOutput >= AInput, AOutput < AInput + N) and drop the now-unused
LSrcAddr/LDstAddr NativeUInt locals. Pointer relational comparison is the same
unsigned address comparison the NativeUInt casts performed.
Left untouched: the alignment fast-path tests in ClpBinaryPrimitives.Swap*,
ClpByteUtilities.Xor/XorTo, and TAesEngineX86.AllocAlignedKeys. These genuinely
need the address as an integer for `and`-masking, so their (harmless) hints
remain - no type alias, no suppression.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.