Signed-off-by: Holger Levsen <holger@layer-acht.org>

Signed-off-by: Holger Levsen <holger@layer-acht.org>

This reverts commit 9abb0e66.

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Changes-By: lintian-brush
Fixes: lintian: debian-changelog-has-wrong-day-of-week
See-also: https://lintian.debian.org/tags/debian-changelog-has-wrong-day-of-week.html

Nis Martensen authored Aug 26, 2020 and Holger Levsen committed Jan 18, 2021

Since commit c3900457

 different dicts
were used for iterating and modifying, so we never modify the dicts we
are currently iterating over. However, the commit was incomplete because
it was longer guaranteed that each key is still part of the dict we are
trying to modify, but no checks were added.
The result are possible crashes such as the one reported in 968842:

    Traceback (most recent call last):
      File "/usr/sbin/piuparts", line 3267, in <module>
        main()
      File "/usr/sbin/piuparts", line 3252, in main
        process_packages(package_list)
      File "/usr/sbin/piuparts", line 3173, in process_packages
        if not install_purge_test(chroot, chroot_state,
      File "/usr/sbin/piuparts", line 2414, in install_purge_test
        return check_results(chroot, chroot_state, file_owners, deps_info=deps_info)
      File "/usr/sbin/piuparts", line 2235, in check_results
        (new, removed, modified) = diff_meta_data(reference_info, current_info)
      File "/usr/sbin/piuparts", line 2041, in diff_meta_data
        del tree2_c[name]
    KeyError: '/var/spool/news/in.coming/tmp/'

Such KeyErrors can be prevented by checking if a key exists in the dict
before trying to drop it. Adding explicit checks is simpler than
adding a try..except clause here because there are already if clauses
that we can adapt.

Closes: #968842
Signed-off-by: Holger Levsen <holger@layer-acht.org>

Nis Martensen authored Sep 06, 2020 and Holger Levsen committed Jan 18, 2021

This should fix the crash reported in
https://alioth-lists.debian.net/pipermail/piuparts-devel/2020-September/009165.html



Signed-off-by: Holger Levsen <holger@layer-acht.org>

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Nis Martensen authored Dec 28, 2020 and Holger Levsen committed Jan 18, 2021

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Nis Martensen authored Dec 27, 2020 and Holger Levsen committed Jan 18, 2021

The existing must-succeed test protects against potential bugs that make
piuparts always fail. Other potential bugs might make piuparts always
succeed, so let's add a must-fail test to cover this case, too.

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Make the summary generation atomic

See merge request !28

Having Depends last confuses apt 2.0.0.

Reported-By: Ferenc Wágner (https://alioth-lists.debian.net/pipermail/piuparts-devel/2020-March/009069.html)

This file, which tracks the current dselect download method, is now handled by
dselect (>= 1.20.0~), but it is also created by debootstrap while it is doing
its apt/dselect setup.

Closes: #953756.

With the performance improvement of querying the BTS in bulk, piuparts-analyze
now spends most of its time waiting for busy sections. Let's wait a bit less.

The BTS SOAP interface struggles when we do a lot of package-based queries; We
even get throttled (and get ConnectionErrors) when doing the queries for a usual
piuparts-analyze run on pejacevic.

Querying the BTS by bug numbers instead doesn't hit the throttle, even when
getting all bugs usertagged for piuparts. This bulk query makes the full
piuparts-analyze run go way faster.

fixes:

    ERROR processing fail/<package>.log
    <class 'TypeError'>

See !10

See merge request !15

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Should fix:

Traceback (most recent call last):
  File "/srv/piuparts.debian.org/share/piuparts/piuparts-report", line 1911, in <module>
    main()
  File "/srv/piuparts.debian.org/share/piuparts/piuparts-report", line 1851, in main
    section.generate_output(output_directory, section_names, problem_list, web_host)
  File "/srv/piuparts.debian.org/share/piuparts/piuparts-report", line 1660, in generate_output
    self.generate_html()
  File "/srv/piuparts.debian.org/share/piuparts/piuparts-report", line 1587, in generate_html
    failures = dwke_get_failures(self._binary_db, self._problem_list)
  File "/srv/piuparts.debian.org/share/piuparts/piuparts-report", line 1761, in dwke_get_failures
    add_cnt = make_kprs(logdict, kprdict, problem_list)
  File "/srv/piuparts.debian.org/lib/python3/dist-packages/piupartslib/dwke.py", line 252, in make_kprs
    logbody = lb.read()
  File "/usr/lib/python3.7/codecs.py", line 322, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xf1 in position 26184: invalid continuation byte

Move __pycache__ directory removal from debian/rules to the `clean`
Makefile target. The change should have no effect on the generated
packages; it just makes `make clean` more complete and debian/rules
simpler.

Instead of using a shell loop for passing the files to `rm`, also use
find's `-delete` option, avoiding any potential issues with weird
filenames.

Since the __pycache__ folders may contain .pyc files, make sure these
are all deleted before.

Piuparts does not have `--list` (without further arguments) or `--info`
options, hence running the autopkgtest currently fails. Replace the two
broken stubs by two working tests:

 1. run `piuparts --version`
 2. create a minimal dummy binary package and run piuparts on it

Using a single-letter package name for the dummy package ensures that
there is no package with the same name in the archive -- according to
policy 5.6.1, all official package names must be at least two characters
long.

Signed-off-by: Holger Levsen <holger@layer-acht.org>

Signed-off-by: Holger Levsen <holger@layer-acht.org>