On 07.10.19 18:28, Thomas Huth wrote:
> On 07/10/2019 15.11, Thomas Huth wrote:
>> On 07/10/2019 14.52, Max Reitz wrote:
>>> On 07.10.19 14:16, Thomas Huth wrote:
>>>> On 04/10/2019 14.44, Max Reitz wrote:
>>>>> On 04.10.19 12:19, Kevin Wolf wrote:
>>>>>> Am 02.10.2019 um 19:47 hat Max Reitz geschrieben:
>>>>>>> On 02.10.19 18:44, Kevin Wolf wrote:
>>>>>>>> Am 02.10.2019 um 13:57 hat Max Reitz geschrieben:
>>>>>>>>> It usually worked fine for me because it’s rather rare that non-block
>>>>>>>>> patches broke the iotests.
>>>>>>>>
>>>>>>>> I disagree. It happened all the time that someone else broke the iotests
>>>>>>>> in master and we needed to fix it up.
>>>>>>>
>>>>>>> OK.
>>>>>>>
>>>>>>> (In my experience, it’s still mostly block patches, only that they tend
>>>>>>> to go through non-block trees.)
>>>>>>
>>>>>> Fair enough, it's usually code that touches block code. I assumed "block
>>>>>> patches" to mean patches that go through one of the block trees and for
>>>>>> which iotests are run before sending a pull request.
>>>>>>
>>>>>> In the end, I don't care what code these patches touched. I do an
>>>>>> innocent git pull, and when I finally see that it's master that breaks
>>>>>> iotests and not my patches on top of it, I'm annoyed.
>>>>>
>>>>> Hm.  Part of my point was that this still happens all the time.
>>>>>
>>>>> Which is why I’d prefer more tests to run in CI than a handful of not
>>>>> very useful ones in make check.
>>>>
>>>> Ok, so let's try to add some more useful test to the "auto" group. Kevin
>>>> mentioned 030, 040 and 041, and I think it should be ok to enable them
>>>> (IIRC the only issue was that they run a little bit longer, but if they
>>>> are very useful, we should include them anyway).
>>>
>>> I agree on those.  (Maybe not 040, but definitely 030 and 041.)
>>>
>>> Maybe one of the issues was the “path too long” thing for Unix sockets?
>>
>> Ah, right. I've applied John's "remove 'linux' from default" patch and
>> added the three iotests to the "auto" group, and indeed, they fail now
>> on cirrus-ci due to the "path too long" socket problem. "We" (royal we,
>> I guess) should likely fix that first...

I’m looking into that, by the way.  I hope a SOCK_DIR that defaults to
$(mktemp -d) will suffice...

> FWIW, 041 also fails on macOS on Travis (which does not have the "path
> too long" issue):
> 
>  https://travis-ci.com/huth/qemu/jobs/242942716#L8415
> 
> ... so we might need to declare this as "linux only" again after John's
> patch gets merged.

:/

Either that, or let make check skip the iotests generally on non-Linux
systems.

Max