[Bug 2077414] Re: gap-guava autopkgtests fail on amd64 with glibc 2.40

[Simon Chopin]

I got to the bottom of it. It turns out not to be a race condition at
all, but rather a case of uninitialized array on the stack. The stack
layout changed a bit in our glibc version.

The fix was simply to initialize the array. It has been posted in a PR
upstream, as well as in a Debian bug (even though they're not affected,
presumably because they don't have the same compilation flags).

In any case, not glibc's fault :)

** Changed in: glibc (Ubuntu)
       Status: New => Invalid

** Changed in: gap-guava (Ubuntu)
       Status: New => Fix Committed

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to glibc in Ubuntu.
https://bugs.launchpad.net/bugs/2077414

Title:
  gap-guava autopkgtests fail on amd64 with glibc 2.40

Status in gap-guava package in Ubuntu:
  Fix Committed
Status in glibc package in Ubuntu:
  Invalid

Bug description:
  The tests fail with the following output:

  ########> Diff in /usr/share/gap/pkg/guava/tst/guava.tst:649
  m
  # Input is:
  p := CodeIsomorphism( C1, C2 );
  # Expected output:
  (2,13,7,10,8,3,5,4,14)(12,15)
  # But found:
  Syntax error: expression expected in /tmp/gaptempfile.XJJWPR:1
  GUAVA_TEMP_VAR :=   &
                     ^
   false
   ########
  ########> Diff in /usr/share/gap/pkg/guava/tst/guava.tst:651
  # Input is:
  C3 := PermutedCode( C1, p );
  # Expected output:
  a linear [15,5,7]5 permuted code
  # But found:
  Error, no method found! For debugging hints type ?Recovery from NoMe\
  \
  thodFound
  Error, no 1st choice method found for `PermutedCode' on 2 arguments
  ########
  ########> Diff in /usr/share/gap/pkg/guava/tst/guava.tst:653
  # Input is:
  C2 = C3;
  # Expected output:
  true
  # But found:
  Error, Variable: 'C3' must have a value
  ########
  #I  Errors detected while testing package guava version 3.19
  #I  using the test file `/usr/share/gap/pkg/guava/tst/guava.tst'
  ########
      2204 ms (103 ms GC) and 425MB allocated for makecheck.tst
  -----------------------------------
  total      2204 ms (103 ms GC) and 425MB allocated
                1 failures in 1 of 1 files

  #I  Errors detected while testing

  I can reproduce the failure using upstream's development branch. Using
  UBSan or AddressSanitizer didn't produce anything useful, however
  using ThreadSanitizer made the tests pass, which makes me think
  there's a race condition somewhere.

  Given that it is the only failure of its kind on all the autopkgtests,
  I'm guessing the fault lies outside of glibc and in either the gap-
  guava code base or the main gap code.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/gap-guava/+bug/2077414/+subscriptions