externalbugtracker-comment-pushing.txt failing on db-devel

[Paul Hummer]

This test is failing on db-devel. I've disabled it for now.

[Björn Tillenius (bjornt)]

On Thu, Mar 05, 2009 at 04:26:07AM -0000, Paul Hummer wrote:
> Public bug reported:
>
> This test is failing on db-devel. I've disabled it for now.

Is is spurious, or is it a real test failure? What is the test failure?
Always include this information when disabling tests, so that it's
easier for the person going to fix it to get an idea of what's wrong.

[Graham Binns (gmb)]

> Is is spurious, or is it a real test failure? What is the test failure?
> Always include this information when disabling tests, so that it's
> easier for the person going to fix it to get an idea of what's wrong.

Here's the failure output from build 120: https://pastebin.canonical.com/14553/

It looks as though the bug messages in the test are being pushed to
the remote bug tracker in the wrong order (i.e. the messages are in
the right sequence and the bodies are as expected but the
remote_comment_ids of the second and third comments are reversed,
which is a bit strange. I can't say at this point whether this is
because we've hit some kind of weird race condition (in which case we
need to guard against it) or whether this is entirely spurious.

I'll dig into this further when I come online properly.

[Graham Binns (gmb)]

This isn't a spurious failure; it's failing consistently locally. I'm trying to isolate the rev the introduced this problem (and hoping it wasn't mine).

[Björn Tillenius (bjornt)]

On Thu, Mar 05, 2009 at 10:52:20AM -0000, Graham Binns wrote:
> This isn't a spurious failure; it's failing consistently locally. I'm
> trying to isolate the rev the introduced this problem (and hoping it
> wasn't mine).

Ok. That makes me wonder, was disabling the tests really appropriate
here? That kind of defeats the purpose of having a testfix mode...

It also makes me wonder, should we make ec2test fire up two instances,
one that tests the branch against devel, and one against db-devel?

[Graham Binns (gmb)]

2009/3/5 Björn Tillenius <email address hidden>:
> Ok. That makes me wonder, was disabling the tests really appropriate
> here? That kind of defeats the purpose of having a testfix mode...

Agreed. A quick branch of db-devel and running the test locally in
isolation showed that it wasn't a spurious failure. Disabling the test
should be the last resort.

> It also makes me wonder, should we make ec2test fire up two instances,
> one that tests the branch against devel, and one against db-devel?

That's an idea, though in this case it might not have helped: the test
doesn't fail when run against devel; the problem is somewhere in
db-devel... I just haven't figured out where yet.

[Graham Binns (gmb)]

The failure was introduced in db-devel r7789, which unfortunately was
a massive merge from mainline, so I don't yet know *what* in that
revision caused the failure.

The failure was subsequently masked by the fact that db-devel was
failing all the time due to the vaccum step in the db restore process.

[Graham Binns (gmb)]

Edwin landed a fix for this on Friday.