Bug#1021085: mksh: ignored test making the build unreproducible
Thorsten Glaser
tg at mirbsd.de
Sat Oct 1 19:57:27 BST 2022
Hi Chris,
> -| FAIL ../../check.t:heredoc-tmpfile-8 (ignored)
> -| Total failed: 1 (1 ignored)
this test is unfortunately fragile. It occasionally fails depending
on system “load” (but not load average), and I haven’t found a way to
make the fragility go away. (If someone does, by all means, tell me.)
>This is because the test summary is printed in README.Debian, and if
>any of the tests are "ignored" this is recorded as well.
OK, ouch.
>To be clear, that mksh includes test info is not necessarily a bug,
>and nor is it necessarily a bug in mksh that some test is failing in
>some weird environment. But the combination of the two is
>unfortunately causing a reproducibility regression and I thought you
>may have more context.
For the test, see above.
I wanted to include the test info somewhere so we can see whether
everything passed; initially, I didn’t fail the build when there
were testsuite failures, but that was in 2008 or so. I also wanted
to be able to gather this information, but nowadays I just use
https://evolvis.org/plugins/scmgit/cgi-bin/gitweb.cgi?p=useful-scripts/useful-scripts.git;a=blob;f=debian/mksh-status;hb=HEAD
which downloads the logs from buildd.d.o and then does the same as
debian/blscan (in the package) does locally to the logs.
I also do a chmod -x on the binaries that build but fail to test,
so we still have them for debugging purposes (especially if, say,
they fail on an emulator but work natively, so we can see that),
and if the build with the system libc fails, abort.
So I don’t, strictly speaking, need this info in the binary pak-
kages any more. At least not for Debian itself; derivates may be
an issue but I don’t actively monitor any at the moment. So, I
*think* I don’t _need_ the entire…
Regression test results:
| FAIL ../../check.t:heredoc-tmpfile-8 (ignored)
| Total failed: 1 (1 ignored)
| Total passed: 578
… block, in either README.Debian in the binary package or the
build log (as the latter has this info inline anyway, even if
it *was* nice-to-have, for quickly seeing this info for all
shell binaries built in one place).
Access to build logs for derivatives is also an issue, but, as
said earlier, currently I’m not monitoring any of these anyway.
I guess I’ll drop this in the next upload, which will be coming
soon since klibc is currently broken on 32-bit platforms; bwh
knows and said he’ll most likely do a release soon.
bye,
//mirabilos
--
Gestern Nacht ist mein IRC-Netzwerk explodiert. Ich hatte nicht damit
gerechnet, darum bin ich blutverschmiert… wer konnte ahnen, daß SIE so
reagier’n… gestern Nacht ist mein IRC-Netzwerk explodiert~~~
(as of 2021-06-15 The MirOS Project temporarily reconvenes on OFTC)
More information about the Reproducible-bugs
mailing list