• Bazel Removal from Debian

    From Olek Wojnar@21:1/5 to All on Sun Feb 19 17:00:01 2023
    Fellow Bazel contributors (and Java Team members),

    I've discussed this with a few of you but I want this to go to the widest possible audience right now. We still have no idea why Bazel Java tests are failing [1] and it seems to indicate an actual problem with Bazel, as opposed to just a flaky test. (It
    would be nice if I'm wrong about that) To save people a click, the failing error is:
    Unrecognized VM option 'UseParallelOldGC'
    (Which has been patched out of the code [2] so I have no idea where it's coming from)

    So, if we cannot fix this, I'll have to remove Bazel from Debian testing before the bookworm release. Meaning Bazel will NOT be in the next stable Debian release.

    1 March 2023 is the last day to make an upload that will definitely make it in before the Hard Freeze.

    So, I'm asking all interested people to please please take a look and see if you can figure out how to fix either the tests or the suspected underlying issue.

    Pull Requests from your own branch in the bazel-bootstrap repository [3] are preferred but I'll take patches or even "you should look at line xxx in this one source file."

    Please reply to this thread with any ideas or suggestions. Hopefully with all our eyes on this we can make this really frustrating bug a bit more shallow.

    Thanks in advance!!

    -Olek

    [1] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/jobs/3862692
    [2] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/blob/master/debian/patches/fix_JDK-16_test_failures.patch
    [3] https://salsa.debian.org/bazel-team/bazel-bootstrap

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Hans-Christoph Steiner@21:1/5 to All on Tue Feb 21 15:40:02 2023
    If the test passes regularly, and just sometimes fails under the same conditions, I would consider disabling it in autopkgtest rather than have it not
    make it into bookworm. The perfect is the enemy of the good.

    .hc

    Olek Wojnar:
    Fellow Bazel contributors (and Java Team members),

    I've discussed this with a few of you but I want this to go to the widest possible audience right now. We still have no idea why Bazel Java tests are failing [1] and it seems to indicate an actual problem with Bazel, as opposed to just a flaky test. (
    It would be nice if I'm wrong about that) To save people a click, the failing error is:
    Unrecognized VM option 'UseParallelOldGC'
    (Which has been patched out of the code [2] so I have no idea where it's coming from)

    So, if we cannot fix this, I'll have to remove Bazel from Debian testing before the bookworm release. Meaning Bazel will NOT be in the next stable Debian release.

    1 March 2023 is the last day to make an upload that will definitely make it in before the Hard Freeze.

    So, I'm asking all interested people to please please take a look and see if you can figure out how to fix either the tests or the suspected underlying issue.

    Pull Requests from your own branch in the bazel-bootstrap repository [3] are preferred but I'll take patches or even "you should look at line xxx in this one source file."

    Please reply to this thread with any ideas or suggestions. Hopefully with all our eyes on this we can make this really frustrating bug a bit more shallow.

    Thanks in advance!!

    -Olek

    [1] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/jobs/3862692
    [2] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/blob/master/debian/patches/fix_JDK-16_test_failures.patch
    [3] https://salsa.debian.org/bazel-team/bazel-bootstrap


    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Emmanuel Bourg@21:1/5 to All on Tue Feb 21 17:10:01 2023
    Le 2023-02-21 16:03, Olek Wojnar a écrit :

    I *have* temporarily disabled the Java tests [1]. As you said, to not
    hold back the package due to a possibly flaky test. But my concern is
    that this is actually an issue with how Bazel uses Java rather than a
    flaky test. Again, I'd be happy to be wrong about that. :)

    This is really just an optimization parameter, I don't think this is
    a critical issue. There is probably a UseParallelOldGC parameter left
    somewhere in the bazel source files. I can't find it on codesearch [1],
    maybe in a zip extracted at build time?

    Emmanuel Bourg

    [1] https://codesearch.debian.net/search?q=UseParallelOldGC+package%3A%5CQbazel-bootstrap%5CE

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Olek Wojnar@21:1/5 to Hans-Christoph Steiner on Tue Feb 21 16:30:01 2023
    Hi hc,

    Thanks for helping to start the conversation!


    On February 21, 2023 2:38:14 PM UTC, Hans-Christoph Steiner <hans@at.or.at> wrote:

    If the test passes regularly, and just sometimes fails under the same conditions, I would consider disabling it in autopkgtest rather than have it not make it into bookworm. The perfect is the enemy of the good.

    Unfortunately, the test fails pretty regularly now. This behavior started when we moved to OpenJDK17. That makes sense since 'UseParallelOldGC' was valid prior to that.

    I *have* temporarily disabled the Java tests [1]. As you said, to not hold back the package due to a possibly flaky test. But my concern is that this is actually an issue with how Bazel uses Java rather than a flaky test. Again, I'd be happy to be wrong
    about that. :)

    -Olek

    [1] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/commit/1c90621d7b9b5d44d01b91dabb04645a20273ca8

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Yun Peng@21:1/5 to olek@debian.org on Tue Feb 21 16:30:01 2023
    My feeling is that this problem is probably fixed in a newer Bazel version,
    so it could be automatically solved if we can package Bazel 5 or 6, but of course, that's not trivial work.

    On Tue, Feb 21, 2023 at 4:04 PM Olek Wojnar <olek@debian.org> wrote:

    Hi hc,

    Thanks for helping to start the conversation!


    On February 21, 2023 2:38:14 PM UTC, Hans-Christoph Steiner <hans@at.or.at> wrote:

    If the test passes regularly, and just sometimes fails under the same conditions, I would consider disabling it in autopkgtest rather than have
    it not make it into bookworm. The perfect is the enemy of the good.

    Unfortunately, the test fails pretty regularly now. This behavior started when we moved to OpenJDK17. That makes sense since 'UseParallelOldGC' was valid prior to that.

    I *have* temporarily disabled the Java tests [1]. As you said, to not hold back the package due to a possibly flaky test. But my concern is that this
    is actually an issue with how Bazel uses Java rather than a flaky test. Again, I'd be happy to be wrong about that. :)

    -Olek

    [1] https://salsa.debian.org/bazel-team/bazel-bootstrap/-/commit/1c90621d7b9b5d44d01b91dabb04645a20273ca8

    --
    Mailing list: https://launchpad.net/~bazel-team
    Post to : bazel-team@lists.launchpad.net
    Unsubscribe : https://launchpad.net/~bazel-team
    More help : https://help.launchpad.net/ListHelp


    <div dir="ltr"><div>My feeling is that this problem is probably fixed in a newer Bazel version, so it could be automatically solved if we can package Bazel 5 or 6, but of course, that&#39;s not trivial work.</div><br><div class="gmail_quote"><div dir="
    ltr" class="gmail_attr">On Tue, Feb 21, 2023 at 4:04 PM Olek Wojnar &lt;<a href="mailto:olek@debian.org">olek@debian.org</a>&gt; wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);
    padding-left:1ex">Hi hc,<br>

    Thanks for helping to start the conversation! <br>


    On February 21, 2023 2:38:14 PM UTC, Hans-Christoph Steiner &lt;<a href="mailto:hans@at.or.at" target="_blank">hans@at.or.at</a>&gt; wrote:<br>
    &gt;<br>
    &gt;If the test passes regularly, and just sometimes fails under the same conditions, I would consider disabling it in autopkgtest rather than have it not make it into bookworm.  The perfect is the enemy of the good.<br>

    Unfortunately, the test fails pretty regularly now. This behavior started when we moved to OpenJDK17. That makes sense since &#39;UseParallelOldGC&#39; was valid prior to that.<br>

    I *have* temporarily disabled the Java tests [1]. As you said, to not hold back the package due to a possibly flaky test. But my concern is that this is actually an issue with how Bazel uses Java rather than a flaky test. Again, I&#39;d be happy to be
    wrong about that. :)<br>

    -Olek<br>

    [1] <a href="https://salsa.debian.org/bazel-team/bazel-bootstrap/-/commit/1c90621d7b9b5d44d01b91dabb04645a20273ca8" rel="noreferrer" target="_blank">https://salsa.debian.org/bazel-team/bazel-bootstrap/-/commit/1c90621d7b9b5d44d01b91dabb04645a20273ca8</a><


    -- <br>
    Mailing list: <a href="https://launchpad.net/~bazel-team" rel="noreferrer" target="_blank">https://launchpad.net/~bazel-team</a><br>
    Post to     : <a href="mailto:bazel-team@lists.launchpad.net" target="_blank">bazel-team@lists.launchpad.net</a><br>
    Unsubscribe : <a href="https://launchpad.net/~bazel-team" rel="noreferrer" target="_blank">https://launchpad.net/~bazel-team</a><br>
    More help   : <a href="https://help.launchpad.net/ListHelp" rel="noreferrer" target="_blank">https://help.launchpad.net/ListHelp</a><br>
    </blockquote></div></div>

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