Fix #22: partial signature on cancel (opt-in)#25
Merged
Conversation
…n it Clarity refactor with zero behavior change. is_unique was already implemented as len(find_all(...)) == 1, throwing away the count. Exposing it gives UniqueSignatureGenerator a clean hook to reuse the search work it was already doing per iteration.
Introduces the opt-in vocabulary for the partial-on-cancel feature. GenerationStatus.UNIQUE preserves existing semantics; GenerationPolicy defaults to strict (raise on cancel) so existing callers see no change.
Adds the result-object fields the generator will populate in the next task. display() prints the partial-format message with the count and intentionally skips Clipboard.set_text for partials so an accidental cancel cannot clobber the user's clipboard.
UniqueSignatureGenerator.generate now returns a GeneratedSignature directly (carrying status + match_count) so partial-on-cancel can thread cleanly. Match count is reused from the per-iteration uniqueness check (rebuilt on count_matches in the previous commit) with zero extra search work. Strict policy (default) preserves the legacy contract. Range generation is unchanged: RangeSignatureGenerator still returns a bare Signature and SignatureMaker.make_signature still wraps it.
Off by default per issue #22 discussion. Users opt in via the form checkbox added in the next task.
Wires the new SigMakerConfig.output_partial_on_cancel field to a new checkbox on the main form (bit 16, default OFF). SigMakerPlugin.run reads it for the CREATE_UNIQUE branch and passes the matching policy to make_signature. Range and XREF paths use the default strict policy and are unaffected.
InstructionWalker had end_ea: int = idaapi.BADADDR as a dataclass field default. Dataclass defaults are evaluated at class-definition time, which under the unit-test mock of idaapi froze end_ea as a MagicMock attribute. Real IDA always supplies a real int, so this was unit-test-only, but it forced TestUniqueSignatureGeneratorPartialOnCancel to mock the whole InstructionWalker rather than the real one. Switch to default_factory=lambda: idaapi.BADADDR so the value is resolved per-instance. Restore the cleaner test which now exercises the real walker.
@OshidaBCF reported testing PR #25 and seeing 'Partial signature (NOT unique, 0 matches)' even when the partial actually matched many places. The cause: SignatureSearcher.find_all polls idaapi_user_canceled inside its scan loop and bails early when set, returning whatever partial count it had so far (often 0). UniqueSignatureGenerator was recording that as last_match_count without checking whether the call had been interrupted. Check idaapi_user_canceled right after count_matches returns. If set, mark last_match_count as None so the partial-on-cancel path shows 'match count unavailable' rather than a falsely-low count.
@OshidaBCF pointed out in issue #22 that 'match count unavailable' is less useful than it could be. The prior fix in e4f4fd4 detected when count_matches was interrupted mid-scan and discarded the unreliable value, but it ALSO discarded the previous iteration's trustworthy count by overwriting last_match_count. This commit changes the guard to a one-way assignment: only update last_match_count when count_matches completed normally. On interruption, last_match_count keeps its prior value (typically the count from the iteration before the user clicked Cancel). The reported number corresponds to a signature one instruction shorter than the partial we emit, so it is an upper bound on the partial's actual match count. That is dramatically more useful than either a fake 0 or a 'match count unavailable' placeholder.
mahmoudimus
added a commit
that referenced
this pull request
May 27, 2026
* docs: add CHANGELOG.md covering 1.7.0 (#24, #25) Keep a Changelog format. Documents the wait-box cancel UX from issue #18 (PR #24) and the opt-in partial-on-cancel output from issue #22 (PR #25), including the user-visible behavior changes, the new opt-in form controls, and the bug fixes that were folded in along the way. * fix: use US spelling in user-facing cancel message The output line on UserCanceledError was 'Operation cancelled by user' (UK), the only UK-spelled user-visible string left in the plugin. Align with the rest of the codebase which uses 'canceled' throughout (per the comment at line 87). * chore: bump version to 1.7.0 Cuts the 1.7.0 release covering the wait-box cancel UX (#24) and opt-in partial-on-cancel output (#25). See CHANGELOG.md for the full set of user-visible and internal changes.
mahmoudimus
added a commit
that referenced
this pull request
May 27, 2026
Brings in PR #24 (issue #18 cancel), PR #25 (issue #22 partial on cancel), and PR #28 (1.7.0 release). The auto-merge handled src/sigmaker/__init__.py cleanly; tests/unit_test_sigmaker.py had overlapping test-class insertions at the same anchor, resolved by taking main's test file as the base and appending PR #26's three new test classes (TestMinimalFunctionSignatureGenerator, TestActionEnumAddsFunctionSig, TestGeneratedSignatureOrdering) before the if __name__ block.
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.
Closes #22.
Background
From @OshidaBCF on issue thread #22:
Today, when a unique-signature search is cancelled, the in-progress signature lives only as a local variable inside
UniqueSignatureGenerator.generateand gets thrown away. The user seesOperation cancelled by userand nothing else. This PR keeps the partial around when the user opts in.Algorithm note
The generator already calls
is_unique(...)on every iteration, implemented aslen(find_all(...)) == 1. That is, every byte we append, we do a full database scan and then throw the count away. So the match count we want to show on cancel was already being computed every iteration in the loop above; we just weren't keeping it.I expose the count via a new
SignatureSearcher.count_matches, rewriteis_uniqueon top of it, and have the generator hold onto the last value it observed. On cancel, the most recent count is exactly the right number for the bytes currently in the partial signature, by construction (the cancel check is at the top of the loop,count_matchesis at the bottom). Zero extra search work.Net behavior
Operation cancelled by user. No change from current behavior.Partial signature (NOT unique, 47 matches) for 0x140001234: E8 ?? ?? ?? ?? 45 33 F6 .... The clipboard is not touched, so an accidental cancel cannot clobber what you had copied before.Operation cancelled by user, so no zero-byte garbage printed.Verification
tests/unit_test_sigmaker.py)idapro-tests(9.0/9.1)idapro-tests-9.2Zero regressions. New test coverage:
TestSignatureSearcherCountMatches:count_matchesequalslen(find_all(...));is_uniquestays True only at exactly 1.TestGenerationStatusAndPolicy: enum members and policy classmethods.TestGeneratedSignatureDisplay: UNIQUE prints and writes clipboard; PARTIAL_ON_CANCEL prints with match count and does not write clipboard; partials never recompute the count.TestUniqueSignatureGeneratorPartialOnCancel: strict cancel raises; permissive cancel returns a partial with the last-observed match count; permissive + unique returns UNIQUE; permissive cancel before any byte appended still raises.TestSigMakerConfigDefaults: new field defaults False.