Re: migration to htslib ecosystem 1.18

Étienne Mollier <[email protected]> Wed, 8 Nov 2023 16:37:44 +0100
Newsgroups	gmane.linux.debian.devel.medical
Message-ID	<ZUurSBePEVngXU-n@fusion>
Hi Andreas,

Andreas Tille, on 2023-11-08:
> No problem.  I also dived into the rdepends of libhts3 and stubled
> upon kallisto which has a new upstream version.  I tried to upgrade
> to this but there is a cmake issue[1]:
> 
> CMake Error at src/CMakeLists.txt:20 (ExternalProject_Get_Property):
>   Unknown CMake command "ExternalProject_Get_Property".
> -- Configuring incomplete, errors occurred!
> 
> I guess it is connected to the internal code copy of bifrost which
> I just packaged[2] (will try a short autopkgtest before uploading
> to new)

Acknowledged.

> We probably also need to talk to debci people since the upload of
> htslib has triggered a failing test for jellyfish on arm64[3].  When
> looking at the log[4] it ends with
> 
> ...
> 182s [ RUN      ] CooperativePool/CooperativePoolTest.Ints/9
> 10051s autopkgtest [06:02:56]: ERROR: timed out on command "su -s /bin/bash debci -c set -e; exec /tmp/autopkgtest-lxc.xayqrapq/downtmp/wrapper.sh --artifacts=/tmp/autopkgtest-lxc.xayqrapq/downtmp/run-unit-test-artifacts --chdir=/tmp/autopkgtest-lxc.xayqrapq/downtmp/build.N8X/src --env=DEB_BUILD_OPTIONS=parallel=4 --env=DEBIAN_FRONTEND=noninteractive --env=LANG=C.UTF-8 --unset-env=LANGUAGE --unset-env=LC_ADDRESS --unset-env=LC_ALL --unset-env=LC_COLLATE --unset-env=LC_CTYPE --unset-env=LC_IDENTIFICATION --unset-env=LC_MEASUREMENT --unset-env=LC_MESSAGES --unset-env=LC_MONETARY --unset-env=LC_NAME --unset-env=LC_NUMERIC --unset-env=LC_PAPER --unset-env=LC_TELEPHONE --unset-env=LC_TIME --script-pid-file=/tmp/autopkgtest_script_pid --source-profile --stderr=/tmp/autopkgtest-lxc.xayqrapq/downtmp/run-unit-test-stderr --stdout=/tmp/autopkgtest-lxc.xayqrapq/downtmp/run-unit-test-stdout --tmp=/tmp/autopkgtest-lxc.xayqrapq/downtmp/autopkgtest_tmp --make-executable=/tmp/autopkgtest-lxc.xayqrapq/downtmp/build.N8X/src/debian/tests/run-unit-test -- /tmp/autopkgtest-lxc.xayqrapq/downtmp/build.N8X/src/debian/tests/run-unit-test" 
> 
> 
> These timeout thingies are really hard to debug.  Maybe we should simply
> shorten the jellyfish test suite to avoid this.  For instance droping
> the `/usr/lib/jellyfish/bin/test_all` call at the end makes the test
> more sensible (also in terms of not stressing debci infrastructure to
> much?
> 
> What do you think?

I did not anticipate that issue with the experimental upload.
Drop of the affected would be the fastest way of "correcting"
the issue.  I'm a bit bugged that this affects the arm64
platform only and not the 32-bit variants, but perhaps the test
suite is already disabled for them.

My "go to" mitigation for these sort of weird issues on arm
platforms would be to rebuild with option -fsigned-char, to
ensure this is not a case of char types being used as a short
integer assumed to be signed.  I'm having a look at these now;
there is also the risk that the test enters the infinite loop
unpredictably.

> Kind regards
>     Andreas.
> 
> 
> [1] https://salsa.debian.org/med-team/kallisto/-/jobs/4901979#L1704
> [2] https://salsa.debian.org/med-team/bifrost
> [3] https://tracker.debian.org/pkg/htslib
> [4] https://ci.debian.net/data/autopkgtest/testing/arm64/j/jellyfish/39674223/log.gz

Have a nice day,  :)
-- 
  .''`.  Étienne Mollier <[email protected]>
 : :' :  gpg: 8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
 `. `'   sent from /dev/pts/2, please excuse my verbosity
   `-    on air: Their Dogs Were Astronauts - Human Resistance
signature.asc (application/pgp-signature, 833 B)