Bug 891813 – dev-libs/castxml-0.5.1 fails tests (LIBARCHIVE-SYSTEM)

• Product: Gentoo Linux
• Status: CONFIRMED ---

[Description Agostino Sarubbo 2023-01-23 07:21:28 UTC]

https://blogs.gentoo.org/ago/2020/07/04/gentoo-tinderbox/

Issue: dev-libs/castxml-0.5.1 fails tests (LIBARCHIVE-SYSTEM).
Discovered on: amd64 (internal ref: libarchive_tinderbox)

NOTE:
(LIBARCHIVE-SYSTEM) in the summary means that the bug was found on a machine that runs app-alternatives/tar and app-alternatives/cpio with [-gnu,+libarchive] but this bug MAY or MAY NOT BE related to the tar/cpio implementation.

[Comment 1 Agostino Sarubbo 2023-01-23 07:21:30 UTC]

Created attachment 849071 [details]
build.log

build log and emerge --info

[Comment 2 Agostino Sarubbo 2023-01-23 07:21:33 UTC]

Created attachment 849073 [details]
1-LastTest.log.xz

1-LastTest.log (compressed because it exceeds attachment limit, use 'xzless' to read it)

[Comment 3 Thomas Beierlein 2023-02-03 18:03:31 UTC]

Seems not related to the LIBATCHIVE-SYSTEM. Fails also with app-alternatives/tar with [+gnu, -libarchive].

It is weird that it now fails, as I fixed the test suite just 10 days before.

[Comment 4 Thomas Beierlein 2023-02-04 17:15:29 UTC]

As it turns out the bump from clang-15.06 to clang-15.0.7 is the main culprit for the failing tests. 

There seems to be two different factors

- For some of the tests the expected output (checked by the test suite) changed so that it gets not recognized as valid. There seems to be only a minority of the failing  tests with that reason.

- For a quite large number of tests castxml just fails to run with something like 
"unsupported option '-fPIE' for target 'i386-pc-windows-msvc'" or similar.


It seems questionable to keep the tests in place if they are so brittle with regard to minor changes of the underlying clang version.

[Comment 5 Thomas Beierlein 2023-02-20 18:20:11 UTC]

*** Bug 895394 has been marked as a duplicate of this bug. ***

[Comment 6 Thomas Beierlein 2023-08-05 12:04:26 UTC]

Did some further test with clang-15.0.7 on a different distribution (Linux MINT). All Tests runs without problems there.

So it seems to be the 'gentoo way' we use in the clang config files.

Further tests showed that the mayor of the failed tests (around 500) are a result of gentoo-hardened.cfg especially the '-fPIE' flag. Further 10 failed tests are a result of including 'maybe-stddefs.h'.