[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: protozero test failure



On 02-09-15 15:37, Jochen Topf wrote:
> On Mi, Sep 02, 2015 at 03:27:15 +0200, Sebastiaan Couwenberg wrote:
>> On 02-09-15 15:04, Jochen Topf wrote:
>>> I saw you disabled the tests in protozero? Any reason for that? It seems to me
>>> 1.1.0-2 built fine for most architectures and the tests are there for a reason.
>>> (Or, if you remove the tests anyway, you can drop quite a lot of build
>>> dependencies.)
>>>
>>> If I read the logs right only armfh failed for -2 and not for -3 and that Bus
>>> error looks bad. I don't really understand whats different about armfh to
>>> even begin to figure out what could have went wrong here.
>>
>> And that's exactly the reason why I ignore test failures (not disable
>> the tests entirely). The build failure on armhf would prevent testing
>> migration of libosmium that built successfully on armhf before but
>> cannot be built without protozero anymore.
>>
>> I inspect the build logs before upload, and most often after upload too.
>> So any issues uncovered by the tests will get noticed, it just won't
>> prevent testing migration anymore because of missing builds on some of
>> the architectures.
>>
>> My take on test failures on ports is to fix what I can to get tests to
>> succeed, and I ignore failures on some of the less well supported ports.
>>
>> We've had those Bus errors for sfgcal on mips too, it's very likely
>> caused by the ongoing GCC 5 transitions through which the slower ports
>> are out of sync. Those ports also suffer from general performance in the
>> case of mips & mipsel (they'll likely be dropped after stretch), or
>> missing FPU support in the case of armel.
> 
> Okay. Makes sense. Unfortunately this way we wont notice if new test failures
> creep in in new versions. Would it be possible to ignore tests only on some
> architectures?

Ignoring test failures on the problematic architectures is a better
option yes, but tends to require a couple of uploads to get the list
complete for the slow and unofficial ports.

Always ignoring test failures is a two line change, ignoring on some
ports only is a 6 line change, so the former tends to be the first choice.

I've now updated protozero to only ignore test failures on armhf because
of the repeated Bus errors that are unlikely issues with the package or
the code itself.

I'll likewise update the osmium family that ignore test failures just
the same. I'd also like to encourage you to commit the changes you'd
like to see in your packages yourself, besides being an upstream
developer you're also a member of the team.

Kind Regards,

Bas

-- 
 GPG Key ID: 4096R/6750F10AE88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146  50D1 6750 F10A E88D 4AF1


Reply to: