Re: master bf7041a: Centralize subprocess creation in a single function.

[Lars Ingebrigtsen]

Paul Eggert <eggert@cs.ucla.edu> writes:

> By the way, after my recent cleanups for the Gnulib merge I'm still
> running into problems running parallel 'make check' on Fedora 33 and
> on Ubuntu 20.10. Plain 'make check' works, but 'make -j4 check'
> hangs. 'top' reports:
>
>     PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+
>     COMMAND
> 1858899 eggert    20   0  373536  33016  20968 R 100.0   0.4   2:18.83
> emacs
> 1858163 eggert    20   0  440044  99836  21144 R  99.7   1.2   2:21.01
> emacs
> 1860091 eggert    20   0  378844  38580  21200 R  99.7   0.5   2:11.89
> emacs
> 1862421 eggert    20   0  489052 142108  21568 R  99.0   1.7   0:17.66
> emacs

I'm not getting a hang, but I'm getting four failures on Fedora 32:

4 unexpected results:
   FAILED  mml-secure-en-decrypt-2
   FAILED  mml-secure-en-decrypt-3
   FAILED  mml-secure-select-preferred-keys-4
   FAILED  mml-secure-sign-verify-1

I'll have a look.

> 1859215 1859215 1859215 ?        00:00:00   gpg-agent
>
> I suspect the subprocess creation changes mentioned in this thread's
> subject line, though I haven't investigated further. Plus, I'm not
> sure why tests would be running GPG connected to my gpg-agent - isn't
> that going a bit far for 'make check'?

The tests start a local gpg-agent, but are supposed to kill them off
again.  I see that no longer works, but I've pushed a fix for that now.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no