[Development] CI broken again?

[Thiago Macieira]

Em qui 06 fev 2014, às 16:01:05, Mandeep Sandhu escreveu:
> It can't be related to my change as the same tests pass 100% of the time on
> my local setup (Ubuntu 13.10) and moreover my changes do not kick-in if the
> name server is not specified (which is the case with all the failing test
> cases).

That doesn't mean anything. The difference might be the type of DNS servers 
available. For example, I wrote a QDnsLookup test that tried to query a record 
containing multiple RR types. It passed on my machine. But on the CI farm, the 
DNS server is "too nice" and sends more information than the DNS request 
really asked for, causing the test to fail.

It's not a CI farm problem. It's a problem with the test I wrote.

I don't know what your change caused of problems. But if the tst_QDnsLookup 
failure happened four times exactly the same way when that change was staged, 
I'm calling that change the culprit. Maybe your code doesn't have a bug and 
this is a problem in the server, maybe it does have a bug you haven't thought 
of, but either way your change causes the test to fail in the farm. You need 
to modify the test so it doesn't fail.

> The latest test failures seemed to be due to timeout rather than anything
> else. The 'waitForDone()' function is exiting with a failure due to DNS
> lookup timeout. The current timeout is set to 10 secs. This is quite large,
> but still, could there be a network issue with the CI setup?

Yes, but it's unlikely to be the case. The only times that tst_QDnsLookup has 
failed in the past 6 months are when your change was staged. The correlation 
factor is way too high.

-- 
Thiago Macieira - thiago.macieira (AT) intel.com
  Software Architect - Intel Open Source Technology Center