blackbox-segfault test - remove residual core files

[grooverdan]

travis test failures on disttest showed corefiles left over. May as well remove them in the blackbox-segfault test.

[knet-ci-bot]

Can one of the admins verify this patch?

[jnpkrn]

ok to test

[fabbione]

ok to test

[jnpkrn]

Ah, I remember now!
#237 (comment)

Anyway, if the immediately preceding qb-blackbox run was what caused
the coredump at hand to be generated, your patch is not expected to
help since there's:

[ $? -gt 127 ] && exit 1 || true

and unless either the libtool's wrapper over the actual compiled binary
or the executing shell, assuming POSIX compatibility[*1], is botched,
the script will return earlier than where you put your clean command.

[*1]
http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_08_02:

The exit status of a command that terminated because it received
a signal shall be reported as greater than 128.

So the best solution appears to turn that AND-OR list into
properly conditionalized block where in the segfault case you
first run a command to print out the backtrace into stderr[*2],
then remove the core, then run exit 1.

[*2] see, e.g.,
https://github.com/jnpkrn/clufter/blob/v0.77.1/.travis.yml#L78-L82

[jnpkrn]

Not to forget, thank you so much for pointing out this fishiness
with Travis CI!

[jnpkrn]

Hmm, I start to see where this comes from.

Please, do the core removal right after

  3 # create a normal blackbox
  4 rm -f crash-test-dummy.fdata
  5 ./crash_test_dummy

where it belongs instead.

[jnpkrn]

Pushed by hand (didn't trigger autoclose).
Thanks!

[jnpkrn]

621b743