Bug#749655: autopkgtest: chokes on libreoffice; leaves 2GB copy of source tree behind in output dir

[Antonio Terceiro]

On Wed, Jun 04, 2014 at 11:46:38AM +0200, Martin Pitt wrote:
> > admin at ip-172-31-36-237:/var/lib/debci/tmp$ echo $?
> > 16
> 
> This means that adt-run or the testbed exit()ed by itself through
> bomb(), and it wasn't terminated by a signal. So the atexit handlers
> should run. What's really strange is that there is no error message at
> all. Every bomb() has an error message which should be visible on
> stderr. Your log definitively has stderr, maybe something got cut off
> somehow? Do you still happen to have this in scrollback?

I don't have it in scrollback anymore (and also didn't when I filed the bug),
but see below.

> > admin at ip-172-31-36-237:/var/lib/debci/tmp$ du -shc libreoffice/*
> > 1.9G	libreoffice/apt0-tests-tree
> 
> In my current run the complete copied up tests-tree is 2.7 GB. Thus I
> conclude that the act.tests_tree.copyup() failed somehow (I don't know
> why, that'd require the error message).
> 
> However, I do see one error in the code:
> 
>     act.tests_tree = TestbedPath(...)
>     act.tests_tree.copyup()
>     atexit.register(rmtree, 'tests-tree', act.tests_tree.host)
> 
> That means the cleanup handler is only registered after a successful
> copyup(), and that's what failed. So I need to register it earlier.
> 
> That will fix the leftover test-tree in the result dir, but of course
> it won't fix the cause why the copyup failed. Are you ok with leaving
> this bug for the cleanup failure?
> 
> If you can reproduce the copyup failure, can you watch out for an
> error message?

I just realized that the reason for the failure is at the very bottom of
http://ci.debian.net/data/packages/unstable/amd64/libr/libreoffice/20140526_142155.autopkgtest.log

adt-run: testing package libreoffice version 1:4.2.4-3
adt-run: * <apt:apt0> build not needed
adt-run: testbed failed: sent `copyup /tmp/adt-run.MY54R7/apt0-build/libreoffice-4.2.4/ /var/lib/debci/data/unstable-amd64/autopkgtest/libr/libreoffice/20140526_142155/apt0-tests-tree/', got `timeout', expected `ok...'

I don't know why I missed that the first time. So all in all I think the
fix you already got is all that can be done in autopkgtest. I am already
discussing with DSA to move ci.debian.net to (better) Debian hardware,
which should fix the root cause.

-- 
Antonio Terceiro <terceiro at debian.org>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/autopkgtest-devel/attachments/20140610/c39631e5/attachment.sig>