Test: regression for iccApplyToLink -PCC profile ownership (#1336)

[colourbill-ctrl]

Summary

Adds a deterministic CTest regression for the iccApplyToLink -PCC leak fixed in #1339 (issue #1336). On an upstream branch because it touches maintainer-owned automation paths (.github/scripts/, Build/Cmake/) that a fork PR cannot modify.

What it guards

iccApplyToLink must release every -PCC profile it opened on its error-exit paths (the leak in #1336 was an AddXform-failure early return that skipped the pccList cleanup). The script drives that path with committed test-data profiles:

iccApplyToLink <out> 0 17 1 Issue1336 0.0 1.0 0 1 \
  .github/ci/test-data/fuzz-mntr-ef352586.icc 1 \
  -PCC Testing/ApplyDataFiles/test-profiles/sRGB_D65_MAT.icc \
  .github/ci/test-data/fuzz-abst-a044ff69.icc 1

The abstract profile doesn't connect → AddXform returns icCmmStatBadSpaceLink → the early return. Pre-fix the -PCC profile is stranded (608 B via CIccProfile::ReadBasic); the run executes under LeakSanitizer and the test fails on any leak report. The tool is expected to exit non-zero (it's reporting the chain error), so the pass/fail signal is the absence of a leak, not the tool exit code.

Verification

• Pre-fix (master without Fix: leak of -PCC profiles in iccApplyToLink on error exit paths (#1336) #1339): ctest -R issue-1336 FAILS — SUMMARY: AddressSanitizer: 608 byte(s) leaked.
• Post-fix (with Fix: leak of -PCC profiles in iccApplyToLink on error exit paths (#1336) #1339): PASSES.
• Uses only tracked profiles — no generated-profile fixture required.

Maintainer note — pairs with #1339 (the iccApplyToLink fix). Merge #1339 first, then this PR. This test is red until then by design: branched off master (which lacks the fix), it reproduces the leak it guards against. Hosted upstream (not the fork) because the fork-PR automation gate forbids fork PRs from touching .github/Build/Cmake.

🤖 Generated with Claude Code

[colourbill-ctrl]

Maintainer note — pairs with #1339 (the iccApplyToLink -PCC fix). Merge #1339 first, then this PR. Expected red until then: branched off master (no fix), the issue-1336 CTest reproduces the 608-byte leak and fails. Turns green once #1339 lands and this branch is updated.

[xsscx]

2026-06-14 03:26:28 UTC

[xsscx]

Note: Auto-Merge is On. Merge on your Schedule.

[xsscx]

Job Fail on CTest

2026-06-14 03:35:54 UTC

[FAIL] iccApplyToLink leaked a -PCC profile on the AddXform error path (#1336)
    #7 0x7f480ed191b2 in CIccProfile::ReadBasic(CIccIO*) /__w/iccDEV/iccDEV/IccProfLib/IccProfile.cpp:1363:12
    #10 0x55cd821da860 in main /__w/iccDEV/iccDEV/Tools/CmdLine/IccApplyToLink/iccApplyToLink.cpp:782:23
    #2 0x55cd821da860 in main /__w/iccDEV/iccDEV/Tools/CmdLine/IccApplyToLink/iccApplyToLink.cpp:782:23
    #2 0x55cd821da860 in main /__w/iccDEV/iccDEV/Tools/CmdLine/IccApplyToLink/iccApplyToLink.cpp:782:23
SUMMARY: AddressSanitizer: 608 byte(s) leaked in 11 allocation(s).

[colourbill-ctrl]

The single failing check (CTest ASAN+UBSAN) is the regression behaving as designed: it reports the 608 byte(s) leaked from the iccApplyToLink.cpp:782 AddXform error path because the fix isn't on master yet. All other checks pass.

It goes green as soon as the paired fix #1339 is merged. Suggested order: merge #1339 (fix) → re-run this PR → merge #1340 (test). This split exists only because the fork-PR automation gate blocks fork PRs that touch test/CI paths, so the fix ships from the fork and the test ships here.

[xsscx]

@colourbill-ctrl #1339 Merged but this Job indicates Fail

[xsscx]

2026-06-14 19:59:31 UTC

Auto-Merge is On. Merge on your schedule.

[colourbill-ctrl]

Thanks @xsscx. The failure was stale, not a real leak: this PR's branch predated the #1339 merge, so the last CI run built iccApplyToLink without the fix and the regression correctly caught the (still-present-on-that-branch) leak.

I've merged current master (which now contains #1339 9b65b40) into this branch and pushed — CI will rebuild with the fix in place.

Verified locally (clang-18, ASAN+UBSAN), fix now in tree:

Test #21: iccdev.issue-1336-applytolink-pcc-regression ... Passed  0.07 sec
100% tests passed, 0 tests failed out of 1

Should go green on the re-run.