runsuite_wrapper.pl is erroneously attributing test failures to i#1938

[xdje42]

Describe the bug

The test wrapper runsuite_wrapper.pl apparently hardwires all expected failures as being due to i#1938.

ref: https://github.com/DynamoRIO/drmemory/blob/release_2.6.0/tests/runsuite_wrapper.pl#L281

To Reproduce

Submit a PR and study the ci-clang workflow test output.

Expected behavior

Each expected test failure should be attributed to the proper bug. Otherwise developers have to spend unnecessary time auditing test failures to ensure new bugs aren't missed.

Screenshots or Pasted Text

From the ci-clang workflow here: https://github.com/DynamoRIO/drmemory/pull/2532

2025-01-13T22:50:33.8979853Z ====> FAILURE in drmemory-dbg-32 <====
2025-01-13T22:50:33.8980200Z drmemory-dbg-32: 113 tests passed, **** 11 tests failed, but ignoring 6 for i1938: ****
2025-01-13T22:50:33.8980541Z    selfmod 
2025-01-13T22:50:33.8980685Z    syscalls_unix 
2025-01-13T22:50:33.8980844Z    clone 
2025-01-13T22:50:33.8980991Z    pthread_test 
2025-01-13T22:50:33.8981155Z    (ignore: i#1938)        wrap_cs2bug 
2025-01-13T22:50:33.8981366Z    (ignore: i#1938)        wrap_operators 
2025-01-13T22:50:33.8981584Z    (ignore: i#1938)        pcache-use 
2025-01-13T22:50:33.8981772Z    realloc 
2025-01-13T22:50:33.8981929Z    (ignore: i#1938)        app_suite 
2025-01-13T22:50:33.8982148Z    (ignore: i#1938)        fuzz_threads 
2025-01-13T22:50:33.8982359Z    (ignore: i#1938)        app_suite.pattern

Versions

HEAD as of commit https://github.com/DynamoRIO/drmemory/commit/f62daf1e14b63cf7e4dad02d4e8f440c92d48c8e OS: Test output reports Ubuntu 20.04.6

[derekbruening]

You can see the actual issue in the comments in that file:

https://github.com/DynamoRIO/drmemory/blob/release_2.6.0/tests/runsuite_wrapper.pl#L203

              # TODO i#2342: These are hitting a DR encoding assert.  Maybe we
                # should just drop wrap_ support anyway.  Also xref i#1741.
                'wrap_malloc' => 1,
                'wrap_cs2bug' => 1,
                'wrap_operators' => 1,

#2342 is the culprit for the wrap_ tests in 32-bit debug.

[xdje42]

Re: reading comments in runsuite_wrapper.pl: Indeed but developers should have be required to do all that digging, reading the test output should be sufficient.

[xdje42]

Data point re: #2342: I'm seeing a different failure for wrap_operators: #2535