1. Forum
  2. Usenet
  3. LINUX.DEBIAN.BUGS.DIST

  • Bug#1104495: [R-pkg-team] Bug#1104495: r-cran-httr2: flaky tests on ris

    From Chris Hofstaedtler@21:1/5 to All on Thu May 1 16:10:02 2025
    XPost: linux.debian.devel.release

    Hi,

    Le Thu, May 01, 2025 at 10:42:03AM +0200, Chris Hofstaedtler a écrit :
    Source: r-cran-httr2
    Severity: serious
    Usertags: flaky

    your package's tests fail very often (but not always) on riscv64:

    * Charles Plessy <plessy@debian.org> [250501 10:56]:

    Control: severity -1 normal

    we are struggling hard with bigger issues, and risc64 is not supported >upstream. Also the whole architecture as a popcon score of 29. Please
    allow us to ignore the issue in this release.

    I think you are mostly hurting your own team, as packages will be
    stuck migrating etc.

    Chris

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Charles Plessy@21:1/5 to All on Fri May 2 03:20:01 2025
    XPost: linux.debian.devel.release

    * Charles Plessy <plessy@debian.org> [250501 10:56]:

    Control: severity -1 normal

    Le Thu, May 01, 2025 at 03:56:52PM +0200, Chris Hofstaedtler a écrit :

    I think you are mostly hurting your own team, as packages will be stuck migrating etc.

    Hi Chris,

    thanks for your help in general; if my actions make your work harder, I will stop changing severities.

    This said, please correct me if anything is wrong in the following:

    - setting the severity from serious to normal is not going to block
    testing migration.

    - r-cran-httr2 is in Trixie at the same version as in Sid, does
    not need migration at the moment, and does not block any migration.

    - Nobody knows how to make r-cran-httr2 tests more robust on risc64
    (except of course by disabling them).

    - keeping severity to serious can cause r-cran-httr2 and its reverse-depencies
    to be removed from Trixie.

    Have a nice day,

    Charles

    --
    Charles Plessy Nagahama, Yomitan, Okinawa, Japan
    Debian Med packaging team http://www.debian.org/devel/debian-med Tooting from home https://framapiaf.org/@charles_plessy
    - You do not have my permission to use this email to train an AI -

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Hofstaedtler@21:1/5 to All on Fri May 2 11:50:01 2025
    XPost: linux.debian.devel.release

    Hi Charles,

    * Charles Plessy <plessy@debian.org> [250502 03:15]:
    * Charles Plessy <plessy@debian.org> [250501 10:56]:
    Le Thu, May 01, 2025 at 03:56:52PM +0200, Chris Hofstaedtler a écrit :

    I think you are mostly hurting your own team, as packages will be stuck
    migrating etc.

    thanks for your help in general; if my actions make your work harder, I will >stop changing severities.

    This said, please correct me if anything is wrong in the following:
    [..]

    - r-cran-httr2 is in Trixie at the same version as in Sid, does
    not need migration at the moment, and does not block any migration.

    Yes, but other packages also trigger the r-cran-httr2 test. For
    example, r-cran-base. Which was blocked from migrating for a while
    because of r-cran-httr2 - at least until someone noticed.

    - Nobody knows how to make r-cran-httr2 tests more robust on risc64
    (except of course by disabling them).

    It looks like the tests have an internal timeout, until when they
    expect something to start. I imagine this timeout can be increased.


    I've briefly spoken to Paul here at MiniDebConf Hamburg. It might be
    reasonable to rejectlist r-cran-httr2 on riscv64, but please
    consider increasing the r-cran-httr2 tests internal timeout first.

    Chris

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Charles Plessy@21:1/5 to All on Sun May 4 01:10:01 2025
    XPost: linux.debian.devel.release

    Le Fri, May 02, 2025 at 11:41:34AM +0200, Chris Hofstaedtler a écrit :

    It looks like the tests have an internal timeout, until when they
    expect something to start. I imagine this timeout can be increased.

    Thanks Chris for checking this,

    Indeed it seems that there is a timeout of 30s that seems enough to
    cause the tests to fail around 5% of the times. I also note that
    some of the failures were synchronised on some architecture (like
    failing on the same day in amd64 and arm64), so it may be network
    issues that are not solvable by changing the timeout.

    I've briefly spoken to Paul here at MiniDebConf Hamburg. It might be >reasonable to rejectlist r-cran-httr2 on riscv64, but please consider >increasing the r-cran-httr2 tests internal timeout first.

    Indeed, the failure rate on risc64 is way higher than on other release architectures. The tests run also 10 times slower, which is a lot. I
    am not familiar with risc64, but if the usage pattern of this
    architecture is narrower than amd64 and arm64 and does not include
    scientific computing, it may make sense to just remove all
    team-maintained r-cran-* packages there too, as it is not supported
    upstream. Risc64 users who just want to perform HTTP requests with a
    script language have better alternatives such as Perl or Python.

    Have a nice day,

    Charles

    --
    Charles Plessy Nagahama, Yomitan, Okinawa, Japan
    Debian Med packaging team http://www.debian.org/devel/debian-med Tooting from work, https://fediscience.org/@charles_plessy Tooting from home, https://framapiaf.org/@charles_plessy

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Hofstaedtler@21:1/5 to All on Sun May 4 10:30:01 2025
    XPost: linux.debian.devel.release

    * Charles Plessy <charles@plessy.org> [250504 01:05]:
    Indeed it seems that there is a timeout of 30s that seems enough to
    cause the tests to fail around 5% of the times. I also note that
    some of the failures were synchronised on some architecture (like
    failing on the same day in amd64 and arm64), so it may be network
    issues that are not solvable by changing the timeout.

    30s might be too slow on all archs, on "loaded" runners.

    I've briefly spoken to Paul here at MiniDebConf Hamburg. It might be >>reasonable to rejectlist r-cran-httr2 on riscv64, but please
    consider increasing the r-cran-httr2 tests internal timeout first.

    Indeed, the failure rate on risc64 is way higher than on other release >architectures. The tests run also 10 times slower, which is a lot. I
    am not familiar with risc64, but if the usage pattern of this
    architecture is narrower than amd64 and arm64 and does not include
    scientific computing, it may make sense to just remove all
    team-maintained r-cran-* packages there too, as it is not supported
    upstream. Risc64 users who just want to perform HTTP requests with a
    script language have better alternatives such as Perl or Python.

    The expectation is that riscv64 will get significantly faster
    hardware (in general and in Debian), but probably during forky.

    I can't make the call for you or the riscv porters.

    Chris

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)