[pkg-gnupg-maint] Bug#867783: gnupg mystery agent failure [and 1 more messages]

[Ian Jackson]

Ian Jackson writes ("gnupg mystery agent failure"):
> The following occurs during CI testing of dgit:
> 
> + gpg --list-secret
> gpg: WARNING: unsafe permissions on homedir '/tmp/autopkgtest-virt-lxc.shared.ar7jp_mz/downtmp/autopkgtest_tmp/gnupg'
> gpg: starting migration from earlier GnuPG versions
> gpg: porting secret keys from '/tmp/autopkgtest-virt-lxc.shared.ar7jp_mz/downtmp/autopkgtest_tmp/gnupg/secring.gpg' to gpg-agent
> gpg: key A3DBCBC039B13D8A/A3DBCBC039B13D8A: error sending to agent: No such file or directory
> gpg: error building skey array: No such file or directory
> gpg: key 2AC1E075F5ED1B25/2AC1E075F5ED1B25: error sending to agent: No such file or directory
> gpg: error building skey array: No such file or directory
> gpg: key 14A868BFAC3BD039/14A868BFAC3BD039: error sending to agent: No such file or directory
> gpg: error building skey array: No such file or directory
> gpg: migration succeeded
> + test 2 = 0
> 
> This does not seem to happen in my own tests using adt-run.

FYI in ci.debian.net this has mysteriously stopped happening.
Observe this list:
  https://ci.debian.net/packages/d/dgit/unstable/amd64/

The debci log for the first recent pass, 2017-07-13 20:37:25 UTC, is
here
  https://ci.debian.net/data/packages/unstable/amd64/d/dgit/20170713_201325.log
and contains useful information about "what changed".
Look for these sections:
  Change in test packages for dgit since last test run
  Change in the base system since last run

I searched for the error message and found these:
  https://bugs.alpinelinux.org/issues/5463
  https://bugzilla.redhat.com/show_bug.cgi?id=1236304
  https://github.com/sobolevn/git-secret/issues/46

Nothing in the list of testbed changes looks particularly relevannt.
Note that the versions of dgit (including test suite) and gnupg (and
related packages) are unchanged.

If it's a race, it's one whose timing is altered by the dependencies,
because the dgit test suite has a lot of individual tests which either
all pass or almost all fail.  That is, the data contradicts the
hypothesis that this is a random failure, independent for each gnupg
migration, whose probability was the same in the tests for 2017-07-12
09:22:51 UTC and 2017-07-13 20:37:25 UTC.

Ian.