How to handle armhf time-out on package?

How to handle armhf time-out on package?

[Pjotr Prins]

According to 

  https://hydra.gnu.org/build/2447190/nixlog/1/tail-reload

we are getting a time-out during testing of dlang ldc on armhf. 

I admit 3600s for a test is a bit lengthy ;)

How to handle this. Should I just disable this test specifically for
armhf?

It can be done using a switch

  (string-prefix? "armhf" (%current-system))

Or should I disable the build for armhf altogether?

Pj.



-- 

Re: How to handle armhf time-out on package?

[Mark H Weaver]

Pjotr Prins <pjotr.public12@thebird.nl> writes:

> According to 
>
>   https://hydra.gnu.org/build/2447190/nixlog/1/tail-reload
>
> we are getting a time-out during testing of dlang ldc on armhf. 
>
> I admit 3600s for a test is a bit lengthy ;)
>
> How to handle this. Should I just disable this test specifically for
> armhf?

No.  Please add a 'max-silent-time' property to the package.  See the
'properties' field of our guile-2.2 package in guile.scm for an example.

     Thanks,
       Mark

Re: How to handle armhf time-out on package?

[Ludovic Courtès]

Hello,

Pjotr Prins <pjotr.public12@thebird.nl> skribis:

> According to 
>
>   https://hydra.gnu.org/build/2447190/nixlog/1/tail-reload
>
> we are getting a time-out during testing of dlang ldc on armhf. 
>
> I admit 3600s for a test is a bit lengthy ;)

Indeed, on x86_64, according to
<https://mirror.hydra.gnu.org/log/q31z6ncd448p4qdcc9xd2j0c08bwbwff-ldc-0.17.4>, we get:

--8<---------------cut here---------------start------------->8---
/gnu/store/6g4pfzgmgwpfkwix0p7r6j349hffh2gs-cmake-3.7.2/bin/ctest --force-new-ctest-process 
Test project /tmp/guix-build-ldc-0.17.4.drv-0/build
        Start   1: build-druntime-ldc-unittest
  1/673 Test   #1: build-druntime-ldc-unittest .................   Passed   36.71 sec
        Start   2: build-druntime-test-runner
  2/673 Test   #2: build-druntime-test-runner ..................   Passed    0.34 sec
        Start   3: build-phobos2-ldc-unittest
  3/673 Test   #3: build-phobos2-ldc-unittest ..................   Passed  487.52 sec
--8<---------------cut here---------------end--------------->8---

So given that the first test is 14 times slower on x86_64, we can expect
the 3rd test to last ~8440 seconds, which exceeds the default
‘max-silent-timeout’.

I’ve added a ’max-silent-time’ property as Mark suggests, for 3h.

Thanks,
Ludo’.

timeout option not honoured (formerly how to handle armhf time-out on package?)

[Pjotr Prins]

the fix did not work, see

  https://hydra.gnu.org/build/2513115

est project /tmp/guix-build-ldc-0.17.4.drv-0/build
        Start   1: build-druntime-ldc-unittest
  1/673 Test   #1: build-druntime-ldc-unittest .................   Passed  533.53 sec
        Start   2: build-druntime-test-runner
  2/673 Test   #2: build-druntime-test-runner ..................   Passed    6.06 sec
        Start   3: build-phobos2-ldc-unittest
building of `/gnu/store/4bqm0has81ndhimqg9nn7kysdzdhssy7-ldc-0.17.4.drv' timed out after 3600 seconds of silence
@ build-failed /gnu/store/4bqm0has81ndhimqg9nn7kysdzdhssy7-ldc-0.17.4.drv - timeout

Strangely no effect on the timeout.

CC debbugs so we don't lose it.

On Fri, Feb 16, 2018 at 10:56:32AM +0100, Ludovic Courtès wrote:
> Hello,
> 
> Pjotr Prins <pjotr.public12@thebird.nl> skribis:
> 
> > According to 
> >
> >   https://hydra.gnu.org/build/2447190/nixlog/1/tail-reload
> >
> > we are getting a time-out during testing of dlang ldc on armhf. 
> >
> > I admit 3600s for a test is a bit lengthy ;)
> 
> Indeed, on x86_64, according to
> <https://mirror.hydra.gnu.org/log/q31z6ncd448p4qdcc9xd2j0c08bwbwff-ldc-0.17.4>, we get:
> 
> --8<---------------cut here---------------start------------->8---
> /gnu/store/6g4pfzgmgwpfkwix0p7r6j349hffh2gs-cmake-3.7.2/bin/ctest --force-new-ctest-process 
> Test project /tmp/guix-build-ldc-0.17.4.drv-0/build
>         Start   1: build-druntime-ldc-unittest
>   1/673 Test   #1: build-druntime-ldc-unittest .................   Passed   36.71 sec
>         Start   2: build-druntime-test-runner
>   2/673 Test   #2: build-druntime-test-runner ..................   Passed    0.34 sec
>         Start   3: build-phobos2-ldc-unittest
>   3/673 Test   #3: build-phobos2-ldc-unittest ..................   Passed  487.52 sec
> --8<---------------cut here---------------end--------------->8---
> 
> So given that the first test is 14 times slower on x86_64, we can expect
> the 3rd test to last ~8440 seconds, which exceeds the default
> ‘max-silent-timeout’.
> 
> I’ve added a ’max-silent-time’ property as Mark suggests, for 3h.
> 
> Thanks,
> Ludo’.
> 

--