failing cuirass i686 tests

failing cuirass i686 tests

[Robert Vollmert]

There are a couple of tests that fail (on master, but also the haskell updates
branch), compare

http://ci.guix.gnu.org/eval/6534?status=failed

They all seem to fail with the following:

%%%% Starting test basic  (Writing full log to "basic.log")
marionette is ready
/gnu/store/s3w6pmyi8gxqpq4gd36qzavd2dsgamgb-installed-os-builder:1: FAIL shell and user commands

Any idea what’s going wrong? Anyone working on fixing this? Is there a way to
get at that full log in “basic.log”? How would I run this test locally?

Cheers
Robert

Re: failing cuirass i686 tests

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 791 bytes --]

Robert Vollmert <rob@vllmrt.net> writes:

> There are a couple of tests that fail (on master, but also the haskell updates
> branch), compare
>
> http://ci.guix.gnu.org/eval/6534?status=failed
>
> They all seem to fail with the following:
>
> %%%% Starting test basic  (Writing full log to "basic.log")
> marionette is ready
> /gnu/store/s3w6pmyi8gxqpq4gd36qzavd2dsgamgb-installed-os-builder:1: FAIL shell and user commands
>
> Any idea what’s going wrong? Anyone working on fixing this? Is there a way to
> get at that full log in “basic.log”? How would I run this test locally?

You can run system tests locally like so:

  make check-system TESTS=installed-os

To change build parameters (e.g. to add #:system) you need to edit
build-aux/run-system-tests.scm.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: failing cuirass i686 tests

[Robert Vollmert]

On 23. Jul 2019, at 18:35, Marius Bakke <mbakke@fastmail.com> wrote:
> 
> Robert Vollmert <rob@vllmrt.net> writes:
> 
>> There are a couple of tests that fail (on master, but also the haskell updates
>> branch), compare
>> 
>> http://ci.guix.gnu.org/eval/6534?status=failed
>> 
>> They all seem to fail with the following:
>> 
>> %%%% Starting test basic  (Writing full log to "basic.log")
>> marionette is ready
>> /gnu/store/s3w6pmyi8gxqpq4gd36qzavd2dsgamgb-installed-os-builder:1: FAIL shell and user commands
>> 
>> Any idea what’s going wrong? Anyone working on fixing this? Is there a way to
>> get at that full log in “basic.log”? How would I run this test locally?

Two updates here:

1. I’ve seen a similar failure on x86_64 now:
http://ci.guix.gnu.org/build/1520368/details
So it seems it’s not i686 specific, just fails more frequently then.

2. The failing test "FAIL shell and user commands” runs basic commands
guix/ls/grep/info. Judging by the test output, those do run successfully. The
order of the output seems to be mixed up, but it looks the same as when the tests
pass. My guess is therefore that it’s rather a problem with the testing infrastructure
rather than with that test specifically.

> 
> You can run system tests locally like so:
> 
>  make check-system TESTS=installed-os
> 
> To change build parameters (e.g. to add #:system) you need to edit
> build-aux/run-system-tests.scm.

Thanks! This takes awfully long on my system unfortunately.

Re: failing cuirass i686 tests

[Ludovic Courtès]

Hello,

Robert Vollmert <rob@vllmrt.net> skribis:

> Two updates here:
>
> 1. I’ve seen a similar failure on x86_64 now:
> http://ci.guix.gnu.org/build/1520368/details
> So it seems it’s not i686 specific, just fails more frequently then.

Indeed, see also:

  https://ci.guix.gnu.org/search?query=basic+spec%3Aguix-master

> 2. The failing test "FAIL shell and user commands” runs basic commands
> guix/ls/grep/info. Judging by the test output, those do run successfully. The
> order of the output seems to be mixed up, but it looks the same as when the tests
> pass. My guess is therefore that it’s rather a problem with the testing infrastructure
> rather than with that test specifically.

I found that this is quite reproducible like this:

  make check-system TESTS=basic
  guix build --check /gnu/store/…-basic.drv
  # Try again until it fails.

I added a ‘pk’ and found that

  (marionette-eval (system "…"))

returns 'ready when the test fails, as if it hadn’t received the 'ready
message before (the one that we get when accessing the ‘repl’ field of
<marionette>).  I don’t see how that can happen, and why precisely here
and not in the previous or next test.  Thoughts?

Ludo’.