bug#22251: ghc-memory test suite fails on i686

bug#22251: ghc-memory test suite fails on i686

[Mark H Weaver]

The ghc-memory test suite has failed several times in a row on i686 with
the same errors:

  http://hydra.gnu.org/build/893578

memory
  [...]
  encoding
    Bytes
      [...]
      BASE32
        [...]
        decode-KAT
          [...]
          11:                                         FAIL
            Exception: Bytes: size must be >= 0
    [...]
    ScrubbedBytes
      [...]
      BASE32
        [...]
        decode-KAT
          [...]
          11:                                         FAIL
            Exception: ScrubbedBytes: size must be >= 0

    Mark

bug#22251: ghc-memory test suite fails on i686

[zimoun]

Hi Mark,

This bug is very old.  And probably not relevant anymore.

On Sun, 27 Dec 2015 at 14:35, Mark H Weaver <mhw@netris.org> wrote:
> The ghc-memory test suite has failed several times in a row on i686 with
> the same errors:
>
>   http://hydra.gnu.org/build/893578

This build number is about openldap-2.4.46.i686-linux on
core-updates-core-updates.

With Guix f4450e8ca909530ee3432710b82f97e23acc7fef, the tests suite of
ghc-memory is turn on and the substitute seems available.

        $ guix weather -s i686-linux ghc-memory
        computing 1 package derivations for i686-linux...
        looking for 2 store items on https://ci.guix.gnu.org...
        updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
        https://ci.guix.gnu.org
          100.0% substitutes available (2 out of 2)
          at least 0.9 MiB of nars (compressed)
          3.2 MiB on disk (uncompressed)
          0.130 seconds per request (0.3 seconds in total)
          7.7 requests per second

          at least 1,000 queued builds
              i686-linux: 222 (22.2%)
              armhf-linux: 287 (28.7%)
              aarch64-linux: 316 (31.6%)
              x86_64-linux: 173 (17.3%)
              i586-gnu: 2 (.2%)
          build rate: 125.58 builds per hour
              x86_64-linux: 40.31 builds per hour
              armhf-linux: 29.13 builds per hour
              i686-linux: 33.78 builds per hour
              aarch64-linux: 20.22 builds per hour
              i586-gnu: 6.36 builds per hour

I am in favor to close it if no more information is added.


All the best,
simon

bug#22251: ghc-memory test suite fails on i686

[Mark H Weaver]

Hi Simon,

zimoun <zimon.toutoune@gmail.com> writes:

> This bug is very old.  And probably not relevant anymore.
>
> On Sun, 27 Dec 2015 at 14:35, Mark H Weaver <mhw@netris.org> wrote:
>> The ghc-memory test suite has failed several times in a row on i686 with
>> the same errors:
>>
>>   http://hydra.gnu.org/build/893578
>
> This build number is about openldap-2.4.46.i686-linux on
> core-updates-core-updates.

I don't recognize the web interface that is currently being served at
<http://hydra.gnu.org/build/893578>.  It certainly does not resemble the
page that was there when I filed the bug.  I'm pretty sure that at the
time, it referred to a failed 'ghc-memory' build.

> With Guix f4450e8ca909530ee3432710b82f97e23acc7fef, the tests suite of
> ghc-memory is turn on and the substitute seems available.

Well, as I vaguely recall (possibly mistakenly), ghc-memory sometimes
built successfully back in 2015 on i686-linux as well.  The issue I
reported was that it often failed.

Unfortunately, last I checked, Cuirass still does not provide anything
close to the level of functionality that Hydra previously provided for
us.  So maybe this bug still exists, or maybe it doesn't.  The
information provided by 'guix weather' is too crude to answer that
question.

Anyway, I agree that we should close this bug.  I'll do so now.
I doubt that many GHC users are on i686 systems anyway.

Thanks again for working through these old bugs, Simon.

     Regards,
       Mark

bug#22251: ghc-memory test suite fails on i686

[zimoun]

Hi Mark,

On Fri, 18 Dec 2020 at 17:12, Mark H Weaver <mhw@netris.org> wrote:

> Well, as I vaguely recall (possibly mistakenly), ghc-memory sometimes
> built successfully back in 2015 on i686-linux as well.  The issue I
> reported was that it often failed.

This could happen.  I remember similar hard time with the package
Guitarix and bug#21803.


> Unfortunately, last I checked, Cuirass still does not provide anything
> close to the level of functionality that Hydra previously provided for
> us.  So maybe this bug still exists, or maybe it doesn't.  The
> information provided by 'guix weather' is too crude to answer that
> question.

I agree.  Running on all the public build farms I am aware of, I get:

--8<---------------cut here---------------start------------->8---
guix weather ghc-memory -s i686-linux --substitute-urls="https://ci.guix.gnu.org https://bayfront.guix.gnu.org https://guix.cbaines.net https://guix.tobias.gr https://builds.guix-patches.cbaines.net"
computing 1 package derivations for i686-linux...

[...]

https://ci.guix.gnu.org
  100.0% substitutes available (2 out of 2)

[...]

https://bayfront.guix.gnu.org
  0.0% substitutes available (0 out of 2)

[...]

https://guix.cbaines.net
  100.0% substitutes available (2 out of 2)

[...]

https://guix.tobias.gr
  0.0% substitutes available (0 out of 2)

[...]

https://builds.guix-patches.cbaines.net
  0.0% substitutes available (0 out of 2)
--8<---------------cut here---------------end--------------->8---

However, I am not convinced that bayfront.guix.gnu.org, guix.tobias.gr
and builds.guix-patches.cbaines.net build for i686 since they do not
even have the package ’hello’.

Well, a second data point is still crude.


> Thanks again for working through these old bugs, Simon.

It is a pleasure, or at least a fun mean to be more familiar with
Guix. :-)

All the best,
simon