Hi Marius, Marius Bakke <mba...@fastmail.com> wrote:
> Mark H Weaver <m...@netris.org> writes: > >> This new version of strace failed to build twice in a row on my Thinkpad >> X200 running the Guix system and Linux-libre 5.2.5. I've not yet made a >> third attempt. strace-5.1 has always built successfully on this system. >> >> Specifically, strace-5.2 fails 4 of its tests on my system. See below >> for the relevant entries from tests/test-suite.log. > > Yay, test failures that vary based on hardware or kernel. Heh :) > I found this commit upstream which touches the four failing tests, > though I don't really understand what it does: > > https://github.com/strace/strace/commit/4377e3a1535a0ec3a42da8a1366ad6943f4efa0e > > Mark: can you try the attached patch and see if it makes a difference? > Strace (still) builds for me with this patch, at least. It made no difference, but further investigation revealed that the changes implied by the patch (e.g. changing -a22 to -a20 in the ipc_msgbuf-Xraw test) were not actually applied in practice. Those switches are included in the 'tests/test-suite.log' output, and they are unchanged from before this patch was applied. It turns out that many other files included in the release tarball are generated from 'tests/gen_tests.in' using the 'tests/gen_tests.sh' script, and they are not automatically updated by the normal build process after patching 'tests/gen_tests.in', at least not when all the files have the same timestamp. To be continued ... Thanks, Mark
