Bug#909689: asterisk: autopkgtest regression

[Bernhard Schmidt]

Hi,

> I don't know how much time I'll have to check on this in the next days.
> Raising severity to block testing migration for now.

Okay, I tried to have a look at this and I have to admit I'm a bit confused.

First, I do see a failing unittest in MY autopkgtest qemu instance that
is not visible on
https://ci.debian.net/data/autopkgtest/testing/amd64/a/asterisk/1270861/log.gz

START  /main/uri/ - uri_parse
[test_uri.c:uri_parse:78]: Condition failed: (uri =
ast_uri_parse(scenario[0]))
END    /main/uri/ - uri_parse Time: <1ms Result: FAIL

Neither main/uri.c nor test_uri.c has been changed in the last couple of
years, and the code looks so dead simple that I can't see anything wrong.

In the end it logs

494 Test(s) Executed  493 Passed  1 Failed

Then the test executes

$asterisk -rx "test generate results txt $results_file"

this is where asterisk is actually segfaulting (not during the
testsuite, but when collecting the results). I haven't managed to pull
the backtrace from the autopkgtest worker yet.

asterisk  8990 21.9  9.3 942052 94380 ?        SLsl 21:49   0:40
/usr/sbin/asterisk -U asterisk -C
/tmp/autopkgtest.Be0rXk/build.Ign/real-tree/debian/tests/testmods/output/asterisk.conf
-U asterisk
+ debian/tests/testmods/asterisk -rx 'test generate results txt
/tmp/autopkgtest.Be0rXk/build.Ign/real-tree/debian/tests/testmods/output/results.txt'
asterisk  8990 21.9  0.0      0     0 ?        Zsl  21:49   0:40
[asterisk] <defunct>

The official Asterisk Jenkins jobs seems to use "test generate results
xml", which

a) does not segfault
b) shows slightly different numbers

<testsuites>
        <testsuite errors="0" time="181.78" tests="495"
name="AsteriskUnitTests">
[...]
                <testcase time="0.0" classname="main.uri" name="uri_parse">
                        <failure><![CDATA[
[test_uri.c:uri_parse:78]: Condition failed: (uri =
ast_uri_parse(scenario[0]))

                ]]></failure>
                </testcase>

So it says 0 errors and one more test than on the CLI output, but does
list the failing unittest.

Tzafrir, any idea?

Bernhard