On my PR 19077 which changes Python/ceval.c, test_gdb fails on Travis CI with Python compiled with clang -Og. The -Og optimization level is a compromise between performance and the ability to debug Python. The problem is that gdb fails to retrieve some information and so test_gdb fails. I proposed bpo-38350 "./configure --with-pydebug should use -O0 rather than -Og", but the status quo is to continue to use -Og by default. See examples of test_gdb failures from PR 19077 below. I propose to skip a test if one of the follow pattern is found in gdb output: * '<optimized out>', * '(frame information optimized out)', * 'Unable to read information on python frame', ====================================================================== FAIL: test_basic_command (test.test_gdb.PyListTests) Verify that the "py-list" command works ---------------------------------------------------------------------- (...) AssertionError: (...) 'Breakpoint 1 at 0x5aabf1: file Python/bltinmodule.c, line 1173. [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". Breakpoint 1, builtin_id (self=, v=42) at Python/bltinmodule.c:1173 1173\t PyObject *id = PyLong_FromVoidPtr(v); Unable to read information on python frame ' did not end with ' 5 6 def bar(a, b, c): 7 baz(a, b, c) 8 9 def baz(*args): >10 id(42) 11 12 foo(1, 2, 3) ' ====================================================================== FAIL: test_bt (test.test_gdb.PyBtTests) Verify that the "py-bt" command works ---------------------------------------------------------------------- (...) AssertionError: 'Breakpoint 1 at 0x5aabf1: file Python/bltinmodule.c, line 1173. [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". Breakpoint 1, builtin_id (self=, v=42) at Python/bltinmodule.c:1173 1173\t PyObject *id = PyLong_FromVoidPtr(v); Traceback (most recent call first): <built-in method id of module object at remote 0x7ffff7f87050> (frame information optimized out) File "/home/travis/build/python/cpython/Lib/test/gdb_sample.py", line 7, in bar baz(a, b, c) File "/home/travis/build/python/cpython/Lib/test/gdb_sample.py", line 4, in foo bar(a, b, c) (frame information optimized out) ' did not match '^.* Traceback \\(most recent call first\\): <built-in method id of module object .*> File ".*gdb_sample.py", line 10, in baz id\\(42\\) File ".*gdb_sample.py", line 7, in bar baz\\(a, b, c\\) File ".*gdb_sample.py", line 4, in foo bar\\(a, b, c\\) File ".*gdb_sample.py", line 12, in <module> foo\\(1, 2, 3\\) '

 New changeset 7bf069b6110278102c8f4719975a5eb5a5af25f9 by Victor Stinner in branch 'master': bpo-40019: Skip test_gdb if Python was optimized (GH-19081) https://github.com/python/cpython/commit/7bf069b6110278102c8f4719975a5eb5a5af25f9 

The fix works as expected: test_gdb passed on PR 19077. Well, likely because a few tests have been skipped. I don't think that in test_gdb can do better than skipping the test if gdb fails to read debug information.

We have this issue in Fedora Rawhide. test_gdb of Python 3.8 fails on s390x and armv7hl architectures with GCC 10: https://bugzilla.redhat.com/show_bug.cgi?id=1818857 I tested manually that my change fix test_gdb: tests are skipped as expected: https://src.fedoraproject.org/rpms/python3/pull-request/182 So I backported the change to 3.7 and 3.8. This change only fix test_gdb. To get a fully working python-gdb.py, IMHO the best is to disable all compiler optimization using -O0 optimization level.

 New changeset 4ced9a7611ddfd923bd8f72aa61121d0e5aeb8fc by Miss Islington (bot) in branch '3.8': bpo-40019: Skip test_gdb if Python was optimized (GH-19081) https://github.com/python/cpython/commit/4ced9a7611ddfd923bd8f72aa61121d0e5aeb8fc 

 New changeset a764a1cc663708361300cdf88fcf697633142319 by Miss Islington (bot) in branch '3.7': bpo-40019: Skip test_gdb if Python was optimized (GH-19081) https://github.com/python/cpython/commit/a764a1cc663708361300cdf88fcf697633142319 

Ok, test_gdb should now be more reliable with various compilers and compiler optimization levels. I chose to not skip a test if '<optimized out>' is found in the gdb output. If it becomes an issue, test_gdb can easily be modified to also be skipped in this case.