fix(#2001): allow cicpTag as a common optional tag in iccPawgReport C5

[colourbill-ctrl]

Fixes #2001.

What was wrong

Check C5 ("Is the profile free of additional tags not required for profile class, other
than allowed optional tags") warned on any profile carrying a cicpTag:

[WARN] C5 ... standard tags outside the local class rule table: 'cicp'

icSigCicpTag was missing from kCommonOptional in PawgReport.cpp.

Why it reached the warning rather than the private-tag bucket

The non-obvious part, and what the added comment records. IsSpecTag() asks
CIccInfo::GetTagSigName, which resolves 'cicp' to "cicpTag" through CIccTagCreator
(IccTagFactory.cpp:138). The name does not begin with "Unknown", so the tag is never
classified as private and falls straight through IsAllowedForClass() to the C5 warning. A
conforming profile was reported as non-conforming.

Not HDR-specific

cicpTag is a v4.4 tag and C5 does not look at the transfer characteristic, so this hits any
profile author who adds one to an ordinary display profile. The regression test is built on a
plain SDR profile for exactly that reason.

Test

Adds .github/ci/regression/pawg-c5-cicp-optional.cpp and the iccdev.pawg-c5-cicp-optional
CTest. It compiles PawgReport.cpp into the test the same way the #1775 compression test does,
then drives DumpPawgReport and reads the C5 line back — C5's verdict is not exposed through
PawgReport.h, and the emitted text is what the issue was reported against.

A/B against one base profile:

case	profile	expected
control	Testing/sRGB_v4_ICC_preference.icc as tracked	C5 OK
subject	the same profile plus a BT.709 cicpTag (1/1/1/0)	C5 OK

Pre-fix only the subject fails, naming 'cicp'; the control passes both ways, which is
what makes a green subject mean something.

The corpus offered nothing to reuse as the subject: of the nine tracked XML carrying a
cicpTag, eight are under Testing/HDR/ and the ninth
(.github/ci/test-data/ub-cicp-colorprimaries-1346.xml) is a deliberately malformed UB fixture
that iccFromXml refuses outright, so it cannot be turned into a profile to assess. Hence the
synthesised subject.

The test writes its own diagnostics to stderr: capturing the report means redirecting
stdout, and there is no portable way to restore it on a CI runner with no /dev/tty or
CONOUT$.

Scoped out on purpose — a related gap this did not fix

The first base profile I tried,
Testing/ApplyDataFiles/test-profiles/sRGB_D65_MAT.icc, was rejected as a control because it
already warns on C5 — for 'c2sp', 's2cp', 'svcn', 'gbd1'. So cicp is not the only
tag missing from these tables.

Those four are deliberately not addressed here: whether they belong in kCommonOptional or
in a per-class rule table is a separate judgement from the one #2001 reports, and folding an
unreviewed table edit into a one-line fix seemed like the wrong trade. Happy to file that
separately if it's wanted.

Pre-flight

lane	result
clang 21.1.3 Release, -Wall -Wextra -Wpedantic -Werror ("strict warnings ENABLED")	grep -cE 'warning:' = 0, errors 0
GCC 15.2.0 in CI's own regression container, strict + -DENABLE_LTO=ON ("strict warnings ENABLED")	grep -cE 'warning:' = 0, errors 0; new CTest passes there too
ASan+UBSan Debug, full ctest	155/157; the 2 failures are pre-existing local noise (tool-coverage CMM profile-chain errors, spectral-tiff-preview missing imagecodecs)
new test under ASAN_OPTIONS=detect_leaks=1	clean — it news a tag and reads a profile, so this was checked rather than assumed

No shell or workflow files changed, so the shellcheck/actionlint pre-flight gates are not
engaged by this diff.

[xsscx]

2026-08-07 10:17:01 UTC

[xsscx]

PR Status

2026-08-07 10:58:29 UTC

1. Rebased on master:HEAD
2. Took BOTH
3. Cross Checks
4. Push
5. Auto-Merge

Success!

Reminder - Please use Stacked PRs https://docs.github.com/en/pull-requests/how-tos/stacked-pull-requests