[Debian-med-packaging] Bug#951944: ariba: FTBFS: dh_auto_test: error: pybuild --test --test-nose -i python{version} -p 3.7 returned exit code 13

[Andreas Tille]

Hi,

On Sun, Feb 23, 2020 at 07:54:59AM +0100, Lucas Nussbaum wrote:
> During a rebuild of all packages in sid, your package failed to build
> on amd64.
> 
> Relevant part (hopefully):
...
> > test _assemble_with_fermilite ... FAIL

I stumbled upon several failures with fermilite in the name of the test.

> > test _assemble_with_spades ... SKIP: Spades assembler is optional and is not configured
> > test _assemble_with_spades handles spades fail ... SKIP: Spades assembler is optional and is not configured
> > test _check_spades_log_file ... Error running SPAdes. Cannot continue. This is the error from the log file /<<PKGBUILDDIR>>/.pybuild/cpython3_3.7_ariba/build/ariba/tests/data/assembly_test_check_spades_log_file.log.bad ...
> > == Error ==  system call for: "['/foo/bar/SPAdes-3.6.0-Linux/bin/spades', '/spam/eggs/K21/configs/config.info']" finished abnormally, err code: -7

This is the only test that is not affected by my considerations.

> > test _run_fermilite ... FAIL
> > test _run_fermilite when it fails ... ok

Here are other examples.

> > --------------------- >> end captured stdout << ----------------------
> > 
> > ----------------------------------------------------------------------
> > Ran 356 tests in 40.055s
> > 
> > FAILED (SKIP=2, failures=26)

I did a local build against fermi-lite 0.1-5 from snapshots[1] which left me with


test _assemble_with_spades ... WARNING: I tried to get the version of spades with: "/usr/bin/spades.py --version" and the output didn't match this regular expression: "SPAdes\s+v([0-9\.]+)"
The following command failed with exit code 1
['python3', '/usr/bin/spades.py', '-t', '1', '--pe1-1', '/home/andreas/debian-maintain/salsa/med-team/build-area/ariba-2.14.4+ds/.pybuild/cpython3_3.7_ariba/build/ariba/tests/data/assembly_test_assemble_with_spa

The output was:

  File "/usr/bin/spades.py", line 2
    SPADESDIR=/usr/lib/spades/bin
              ^
SyntaxError: invalid syntax

FAIL
test _assemble_with_spades handles spades fail ... WARNING: I tried to get the version of spades with: "/usr/bin/spades.py --version" and the output didn't match this regular expression: "SPAdes\s+v([0-9\.]+)"
The following command failed with exit code 1
['python3', '/usr/bin/spades.py', '-t', '1', '--pe1-1', '/home/andreas/debian-maintain/salsa/med-team/build-area/ariba-2.14.4+ds/.pybuild/cpython3_3.7_ariba/build/ariba/tests/data/assembly_test_assemble_with_spa

The output was:

  File "/usr/bin/spades.py", line 2
    SPADESDIR=/usr/lib/spades/bin
              ^
SyntaxError: invalid syntax

ok
test _check_spades_log_file ... Error running SPAdes. Cannot continue. This is the error from the log file /home/andreas/debian-maintain/salsa/med-team/build-area/ariba-2.14.4+ds/.pybuild/cpython3_3.7_ariba/buil
== Error ==  system call for: "['/foo/bar/SPAdes-3.6.0-Linux/bin/spades', '/spam/eggs/K21/configs/config.info']" finished abnormally, err code: -7

ok

...
======================================================================
FAIL: test _assemble_with_spades
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/andreas/debian-maintain/salsa/med-team/build-area/ariba-2.14.4+ds/.pybuild/cpython3_3.7_ariba/build/ariba/tests/assembly_test.py", line 128, in test_assemble_with_spades
    self.assertTrue(a.assembled_ok)
AssertionError: False is not true

----------------------------------------------------------------------
Ran 356 tests in 75.642s

FAILED (failures=1)


I need to inspect the spades issue (may be my local environment is not
totally clean regarding spades!) but it seems that changes in fermi-lite
between 0.1-5 and latest fermi-lite are responsible for most of the test
suite errors.  There might be a chance to "bisect" fermi-lite package
uploads (= checking 0.1-7 and depending from the result 0.1-6 or 0.1-8)
to find out what actual change might have caused the failures.  But
may be Michael or Sascha (both in CC have a more direct idea about the
changes that might affect the test.

Kind regards

    Andreas.

PS: The package is of some specific relevance since it belongs to the
    Python 3.8 transition blockers.


[1] https://snapshot.debian.org/package/fermi-lite/0.1-5/

-- 
http://fam-tille.de