Bug#1112662: gcr4: rare test failure: test-ssh-agent-process.c:80:connect_to_process: 'test->connection' should not be NULL

[Simon McVittie]

Source: gcr4
Version: 4.4.0.1-6
Severity: important
Tags: unreproducible help

In one attempt to run the gcr4 autopkgtests under autopkgtest-virt-qemu, 
I saw this failure:

363s [/usr/libexec/installed-tests/gcr4/test-ssh-agent-preload]: OK
363s [/usr/libexec/installed-tests/gcr4/test-ssh-agent-process]: running...
363s debian/tests/installed-tests: line 9:  2067 Aborted                 timeout -v -k30s 5m $test 2>&1
363s TAP version 14
363s # random seed: R02S042d37aee19998e0da0b03907ee3376a
363s 1..7
363s # Start of ssh-agent tests
363s # Start of process tests
363s **
363s Gcr:ERROR:../../../gcr/test-ssh-agent-process.c:80:connect_to_process: 'test->connection' should not be NULL
363s not ok /ssh-agent/process/connect - Gcr:ERROR:../../../gcr/test-ssh-agent-process.c:80:connect_to_process: 'test->connection' should not be NULL
363s Bail out!
363s [/usr/libexec/installed-tests/gcr4/test-ssh-agent-process]: FAILED exit=134

(This is with a slightly modified debian/tests/installed-tests which 
runs the tests under timeout(1), which I plan to upload soon.)

It does not appear to be a reproducible failure - I've only seen it 
once, and never at build-time (but this test does get run at build time, 
and if the failure happened there, it would cause a FTBFS).

As with #1112659, it is a valid bug but I don't have the spoons or 
domain knowledge to dig into it in exhaustive detail right now, so I'm 
going to mark this test to not be run by default in order to mitigate 
this.

    smcv