DupFileManager, FileMonitor, RenameFile: bind the exception in retry handlers

[johnkattenhorn]

The problem

Each retry loop in StashPluginHelper.py has two exception handlers: a narrow one for connection errors, and a broad one for everything else. Both build the same log message including {e} — but only the broad handler binds the name with as e.

Because the sibling handler binds e as a function local, the narrow handler raises UnboundLocalError while formatting its own log message. That propagates out of the retry loop, so a dropped connection — the exact condition the loop exists to survive — aborts the operation on the first failure and reports the wrong error.

Reproduction, matching the structure in the source:

def retry():
    for i in range(3):
        try:
            raise ConnectionResetError("connection dropped")
        except (ConnectionResetError):
            print(f"retrying: {e}")     # UnboundLocalError
        except Exception as e:
            print(f"retrying: {e}")

Scope

13 sites across the three plugins that vendor StashPluginHelper.py:

File	Lines
plugins/DupFileManager/StashPluginHelper.py	522, 713, 748, 769
plugins/FileMonitor/StashPluginHelper.py	522, 713, 748, 769
plugins/RenameFile/StashPluginHelper.py	522, 713, 748, 769
plugins/DupFileManager/DupFileManager.py	351

Those are the mergeMetadata, addTag, updateScene and destroyScene retry loops in each copy of the helper, plus setTagId_withRetry in DupFileManager.

Every change is the same one-token edit — adding as e to a handler that already interpolated e. No behaviour changes beyond the retry path now working as written.

Testing

Tested against a clean Stash v0.31.1 instance in Docker with a two-scene library, driving the plugin helper through the real plugin entry path (plugin JSON on stdin, server_connection context).

Fault injection: mergeMetadata was made to raise ConnectionResetError on its first two attempts, then succeed — the scenario the retry loop is written for. Same test run against both the current main and this branch:

RESULT[original]: UnboundLocalError: cannot access local variable 'e'
                  where it is not associated with a value   (attempts = 1)
RESULT[patched]:  survived the resets and completed;        attempts = 3

main aborts on the first reset. With the fix, the loop retries and completes as intended.

Also confirmed:

• A normal (non-failing) mergeMetadata between two scenes behaves identically before and after — no regression in the success path.
• AST pass over the whole repository: zero remaining except handlers that reference an unbound e.
• All four modified files compile.

plugins/starIdentifier/star_identifier.py matches this pattern at first glance but is not affected — load_encodings() binds e = Exception(...) before the try, so its handler resolves the name from the enclosing scope. Deliberately left alone.

Note for maintainers

StashPluginHelper.py is vendored into three plugins and the copies have already drifted — DupFileManager's differs from the other two. This fix is applied to all three. Worth considering whether that file wants a single shared home.

LLM-assisted contribution disclosure

Per the repository's contribution policy: this change was prepared with LLM assistance (Claude). The diff has been reviewed by me, the testing described above was carried out and its results are reproduced verbatim, and I take full responsibility for the change and its license compliance.