bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Leo Famulari]

I noticed while building the package added by ng0's patch (below) that
the test suite fails, but the check phase succeeds:

[...]
======================================================================
ERROR: test_clone_with_credentials (test.test_repository.CloneRepositoryTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/test/test_repository.py", line 544, in test_clone_with_credentials
    self._temp_dir, callbacks=pygit2.RemoteCallbacks(credentials=pygit2.UserPass("libgit2", "libgit2")))
  File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/__init__.py", line 255, in clone_repository
    check_error(err)
  File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/errors.py", line 64, in check_error
    raise GitError(message)
_pygit2.GitError: curl error: Couldn't resolve host 'bitbucket.org'


----------------------------------------------------------------------
Ran 262 tests in 5.771s

FAILED (errors=3)
phase `check' succeeded after 8.9 seconds
[...]

Any ideas?

On Sun, Dec 11, 2016 at 06:12:31PM +0000, ng0 wrote:
> * gnu/packages/python.scm (python-pygit2): New variable.
> ---
>  gnu/packages/python.scm | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)
> 
> diff --git a/gnu/packages/python.scm b/gnu/packages/python.scm
> index 2a49a060b..32f83deb4 100644
> --- a/gnu/packages/python.scm
> +++ b/gnu/packages/python.scm
> @@ -3376,6 +3376,36 @@ association studies (GWAS) on extremely large data sets.")
>  (define-public python2-numpy
>    (package-with-python2 python-numpy))
>  
> +(define-public python-pygit2
> +  (package
> +    (name "python-pygit2")
> +    (version "0.24.2")
> +    (source
> +     (origin
> +       (method url-fetch)
> +       (uri (string-append "https://github.com/libgit2/"
> +                           "pygit2/archive/v" version ".tar.gz"))
> +       (file-name (string-append name "-" version ".tar.gz"))
> +       (sha256
> +        (base32
> +         "1j7zkz2bsn4wqmkg1zkxb3r00xx28rkz9jdwcsikcl4ffa298xwa"))))
> +    (build-system python-build-system)
> +    (propagated-inputs
> +     `(("python-six" ,python-six)
> +       ("python-cffi" ,python-cffi)
> +       ("libgit2" ,libgit2)
> +       ("python-tox" ,python-tox)))
> +    (home-page "http://github.com/libgit2/pygit2")
> +    (synopsis "Python bindings for libgit2")
> +    (description
> +     "Pygit2 is a set of Python bindings to the libgit2
> +shared library, libgit2 implements Git plumbing.")
> +    ;; GPL2.0 only, with linking exception.
> +    (license license:gpl2)))
> +
> +(define-public python2-pygit2
> +  (package-with-python2 python-pygit2))
> +
>  (define-public python-pyparsing
>    (package
>      (name "python-pyparsing")
> -- 
> 2.11.0
> 
> 

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 2309 bytes --]

Leo Famulari <leo@famulari.name> writes:

> I noticed while building the package added by ng0's patch (below) that
> the test suite fails, but the check phase succeeds:
>
> [...]
> ======================================================================
> ERROR: test_clone_with_credentials (test.test_repository.CloneRepositoryTest)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/test/test_repository.py", line 544, in test_clone_with_credentials
>     self._temp_dir, callbacks=pygit2.RemoteCallbacks(credentials=pygit2.UserPass("libgit2", "libgit2")))
>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/__init__.py", line 255, in clone_repository
>     check_error(err)
>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/errors.py", line 64, in check_error
>     raise GitError(message)
> _pygit2.GitError: curl error: Couldn't resolve host 'bitbucket.org'
>
>
> ----------------------------------------------------------------------
> Ran 262 tests in 5.771s
>
> FAILED (errors=3)
> phase `check' succeeded after 8.9 seconds
> [...]
>
> Any ideas?

It looks like the 'check' phase ends with 'delete-file-recursively',
which has an unspecified return value, and that eventual failures from
'call-setuppy' are lost.

--8<---------------cut here---------------start------------->8---
(define* (check #:key tests? test-target use-setuptools? #:allow-other-keys)
  "Run the test suite of a given Python package."
  (if tests?
      ;; Running `setup.py test` creates an additional .egg-info directory in
      ;; build/lib in some cases, e.g. if the source is in a sub-directory
      ;; (given with `package_dir`). This will by copied to the output, too,
      ;; so we need to remove.
      (let ((before (find-files "build" "\\.egg-info$" #:directories? #t)))
        (call-setuppy test-target '() use-setuptools?)
        (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
               (inter (lset-difference eqv? after before)))
          (for-each delete-file-recursively inter)))
    #t))
--8<---------------cut here---------------end--------------->8---

Perhaps something like this would work (untested)?


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

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 792 bytes --]

Marius Bakke <mbakke@fastmail.com> writes:

> Perhaps something like this would work (untested)?

Oops, here is the patch I wanted to attach:

--8<---------------cut here---------------start------------->8---
(define* (check #:key tests? test-target use-setuptools? #:allow-other-keys)
  "Run the test suite of a given Python package."
  (if tests?
      (let ((before (find-files "build" "\\.egg-info$" #:directories? #t)))
        (if (call-setuppy test-target '() use-setuptools?)
            (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
                   (inter (lset-difference eqv? after before)))
              (for-each delete-file-recursively inter)
              #t))
        #f)
      #t))
--8<---------------cut here---------------end--------------->8---

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

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 251 bytes --]

On Mon, Dec 12, 2016 at 12:05:57AM +0100, Marius Bakke wrote:
> Marius Bakke <mbakke@fastmail.com> writes:
> 
> > Perhaps something like this would work (untested)?
> 
> Oops, here is the patch I wanted to attach:

Can you make it a patch? :)

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Marius Bakke]

[-- Attachment #1.1: Type: text/plain, Size: 406 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Mon, Dec 12, 2016 at 12:05:57AM +0100, Marius Bakke wrote:
>> Marius Bakke <mbakke@fastmail.com> writes:
>> 
>> > Perhaps something like this would work (untested)?
>> 
>> Oops, here is the patch I wanted to attach:
>
> Can you make it a patch? :)

I'm testing it now on a branch, will report on the results in while. Here
it is in patch form:


[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-build-system-python-Make-sure-return-value-from-chec.patch --]
[-- Type: text/x-patch, Size: 1554 bytes --]

From e069a7f1927ade7a178cb876ee4b4c339a093ba5 Mon Sep 17 00:00:00 2001
From: Marius Bakke <mbakke@fastmail.com>
Date: Mon, 12 Dec 2016 00:25:27 +0100
Subject: [PATCH] build-system/python: Make sure return value from 'check' is
 preserved.

* guix/build/python-build-system.scm (check): Wrap 'call-setuppy' in 'if' so
that it actually fails when the tests fail.
---
 guix/build/python-build-system.scm | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/guix/build/python-build-system.scm b/guix/build/python-build-system.scm
index 3f280b0ac..20eb24edc 100644
--- a/guix/build/python-build-system.scm
+++ b/guix/build/python-build-system.scm
@@ -137,11 +137,13 @@
       ;; (given with `package_dir`). This will by copied to the output, too,
       ;; so we need to remove.
       (let ((before (find-files "build" "\\.egg-info$" #:directories? #t)))
-        (call-setuppy test-target '() use-setuptools?)
-        (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
-               (inter (lset-difference eqv? after before)))
-          (for-each delete-file-recursively inter)))
-    #t))
+        (if (call-setuppy test-target '() use-setuptools?)
+            (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
+                   (inter (lset-difference eqv? after before)))
+              (for-each delete-file-recursively inter)
+              #t))
+        #f)
+      #t))
 
 (define (get-python-version python)
   (let* ((version     (last (string-split python #\-)))
-- 
2.11.0

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Marius Bakke]

[-- Attachment #1.1: Type: text/plain, Size: 1836 bytes --]

Marius Bakke <mbakke@fastmail.com> writes:

> Leo Famulari <leo@famulari.name> writes:
>
>> I noticed while building the package added by ng0's patch (below) that
>> the test suite fails, but the check phase succeeds:
>>
>> [...]
>> ======================================================================
>> ERROR: test_clone_with_credentials (test.test_repository.CloneRepositoryTest)
>> ----------------------------------------------------------------------
>> Traceback (most recent call last):
>>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/test/test_repository.py", line 544, in test_clone_with_credentials
>>     self._temp_dir, callbacks=pygit2.RemoteCallbacks(credentials=pygit2.UserPass("libgit2", "libgit2")))
>>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/__init__.py", line 255, in clone_repository
>>     check_error(err)
>>   File "/tmp/guix-build-python-pygit2-0.24.2.drv-0/pygit2-0.24.2/pygit2/errors.py", line 64, in check_error
>>     raise GitError(message)
>> _pygit2.GitError: curl error: Couldn't resolve host 'bitbucket.org'
>>
>>
>> ----------------------------------------------------------------------
>> Ran 262 tests in 5.771s
>>
>> FAILED (errors=3)
>> phase `check' succeeded after 8.9 seconds
>> [...]
>>
>> Any ideas?
>
> It looks like the 'check' phase ends with 'delete-file-recursively',
> which has an unspecified return value, and that eventual failures from
> 'call-setuppy' are lost.

The good news: the attached patch makes 'check' work as expected. The
bad news is that we have some breakages.

'python-py' fails with:

TypeError: py.test.__dict__ is not a dictionary

Which seems similar to

https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174165144

I tried adding a newer setuptools as input, to no avail. Needs more
investigation.


[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-build-system-python-Make-sure-check-returns-failures.patch --]
[-- Type: text/x-patch, Size: 1648 bytes --]

From 12c4a14b415f26155f85f8ea172ddf696e274855 Mon Sep 17 00:00:00 2001
From: Marius Bakke <mbakke@fastmail.com>
Date: Mon, 12 Dec 2016 00:25:27 +0100
Subject: [PATCH] build-system/python: Make sure 'check' returns failures.

* guix/build/python-build-system.scm (check): Wrap 'call-setuppy' in 'if' so
that it actually fails when the tests fail. Print informational message when
skipped.
---
 guix/build/python-build-system.scm | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/guix/build/python-build-system.scm b/guix/build/python-build-system.scm
index 3f280b0ac..dd07986b9 100644
--- a/guix/build/python-build-system.scm
+++ b/guix/build/python-build-system.scm
@@ -137,11 +137,15 @@
       ;; (given with `package_dir`). This will by copied to the output, too,
       ;; so we need to remove.
       (let ((before (find-files "build" "\\.egg-info$" #:directories? #t)))
-        (call-setuppy test-target '() use-setuptools?)
-        (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
-               (inter (lset-difference eqv? after before)))
-          (for-each delete-file-recursively inter)))
-    #t))
+        (if (call-setuppy test-target '() use-setuptools?)
+            (let* ((after (find-files "build" "\\.egg-info$" #:directories? #t))
+                   (inter (lset-difference eqv? after before)))
+              (for-each delete-file-recursively inter)
+              #t)
+            #f))
+      (begin
+        (format #t "test suite not run~%")
+        #t)))
 
 (define (get-python-version python)
   (let* ((version     (last (string-split python #\-)))
-- 
2.11.0

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 1007 bytes --]

On Mon, Dec 12, 2016 at 08:54:40AM +0100, Marius Bakke wrote:
> The good news: the attached patch makes 'check' work as expected. The
> bad news is that we have some breakages.
> 
> 'python-py' fails with:
> 
> TypeError: py.test.__dict__ is not a dictionary
> 
> Which seems similar to
> 
> https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174165144

Yikes, I had hoped to avoid addressing that Nix issue and the humongous
"fix" for a while longer:

https://github.com/NixOS/nixpkgs/pull/12552

> I tried adding a newer setuptools as input, to no avail. Needs more
> investigation.

Okay, I'm happy to hear that it didn't work ;) My understanding is that
with the new python-build-system, packages are built with the setuptools
that comes with Python, not the external python-setuptools package.

Please let us know what you find.

Hartmut, any thoughts?

In any case, I think it's time to start working on a 'python-updates'
branch, which could include this patch.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Hartmut Goebel]

First of all thanks for spotting this bug.

>> The bad news is that we have some breakages.
>>
>> 'python-py' fails with:
>>
>> TypeError: py.test.__dict__ is not a dictionary
>>
>> Which seems similar to
>>
>> https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174165144

The relevant comment is
https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174196194:
Starting with version 18.4, setuptools will always try to execute a
test-suite (see
https://setuptools.readthedocs.io/en/latest/history.html#id186), which
will fail if there is none.

So the solution is to disable the test-suite for python-py, as there is
no test-suite which can be run via "setup.py test". For testing I added
"python-setuptools" (18.3.1) as native input. This made the "check"
phase run "0 tests" for python2-py and no tests at al for python-py.

(This package includes a test-suite (see tox.ini), but this test-suite
requires py.test, with itself requires python-py. So I suggest to
disable it.)

Our Python (3.5.2) comes with setuptools 20.10.1.

> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
> "fix" for a while longer:
>
> https://github.com/NixOS/nixpkgs/pull/12552

This puill-request is huge, but for setuptools, it comes down that they
updated from 18.2 to 19.4.

-- 
Regards
Hartmut Goebel

| Hartmut Goebel          | h.goebel@crazy-compilers.com               |
| www.crazy-compilers.com | compilers which you thought are impossible |

bug#25177: Test failures don't cause some Python packages to fail [was Re: [PATCH 05/11] gnu: Add python-pygit2.]

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 2202 bytes --]

Hartmut Goebel <h.goebel@crazy-compilers.com> writes:

> First of all thanks for spotting this bug.
>
>>> The bad news is that we have some breakages.
>>>
>>> 'python-py' fails with:
>>>
>>> TypeError: py.test.__dict__ is not a dictionary
>>>
>>> Which seems similar to
>>>
>>> https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174165144
>
> The relevant comment is
> https://github.com/NixOS/nixpkgs/issues/12565#issuecomment-174196194:
> Starting with version 18.4, setuptools will always try to execute a
> test-suite (see
> https://setuptools.readthedocs.io/en/latest/history.html#id186), which
> will fail if there is none.
>
> So the solution is to disable the test-suite for python-py, as there is
> no test-suite which can be run via "setup.py test". For testing I added
> "python-setuptools" (18.3.1) as native input. This made the "check"
> phase run "0 tests" for python2-py and no tests at al for python-py.
>
> (This package includes a test-suite (see tox.ini), but this test-suite
> requires py.test, with itself requires python-py. So I suggest to
> disable it.)

I tried upgrading pytest to 3.0.5 and adding a variant that does not
propagate 'python-py' as input for the 'python-py' package. That got a
bit further, but it does not accept modules from the store!

"ImportError: 'test' module incorrectly imported from '/gnu/store/mwci3544rk71p131mgri5am0sxbz09a6-python-pytest-minimal-3.0.5/lib/python3.5/site-packages'. Expected '/tmp/guix-build-python-py-1.4.31.drv-0/py-1.4.31/py'. Is this module globally installed?"

Disabled tests for now.

> Our Python (3.5.2) comes with setuptools 20.10.1.

Is the 2.7 variant the same version? When adding "python-setuptools" as
native-input, is it guaranteed to override the bundled one?

>> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
>> "fix" for a while longer:
>>
>> https://github.com/NixOS/nixpkgs/pull/12552
>
> This puill-request is huge, but for setuptools, it comes down that they
> updated from 18.2 to 19.4.

Sounds like we're going down the same road. I've started a branch with
the earlier patch and a few other fixes. Is it ok to overwrite the
existing 'python-updates' branch on Savannah?

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

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 722 bytes --]


>>> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
>>> "fix" for a while longer:
>>>
>>> https://github.com/NixOS/nixpkgs/pull/12552
>>
>> This puill-request is huge, but for setuptools, it comes down that they
>> updated from 18.2 to 19.4.
>
> Sounds like we're going down the same road. I've started a branch with
> the earlier patch and a few other fixes. Is it ok to overwrite the
> existing 'python-updates' branch on Savannah?

There are some commits in python-updates that haven't made its way into
any other branches. I created a new branch 'python-tests' for this fix.

Please build and fix as much as possible. I'd like to get this merged
ASAP. @Leo can you start this branch on Hydra?

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 1200 bytes --]

On Tue, Dec 13, 2016 at 09:21:44PM +0100, Marius Bakke wrote:
> 
> >>> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
> >>> "fix" for a while longer:
> >>>
> >>> https://github.com/NixOS/nixpkgs/pull/12552
> >>
> >> This puill-request is huge, but for setuptools, it comes down that they
> >> updated from 18.2 to 19.4.
> >
> > Sounds like we're going down the same road. I've started a branch with
> > the earlier patch and a few other fixes. Is it ok to overwrite the
> > existing 'python-updates' branch on Savannah?
> 
> There are some commits in python-updates that haven't made its way into
> any other branches. I created a new branch 'python-tests' for this fix.
> 
> Please build and fix as much as possible. I'd like to get this merged
> ASAP. @Leo can you start this branch on Hydra?

Several of the commit messages were truncated because of lines that
began with '#:tests?'. The '#' character creates a comment.

I fixed the commit messages with `git rebase`. This means that most of
the commits are now signed by me. I didn't change the commits
themselves.

The evaluation is pending:

https://hydra.gnu.org/jobset/gnu/python-tests

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 1519 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Tue, Dec 13, 2016 at 09:21:44PM +0100, Marius Bakke wrote:
>> 
>> >>> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
>> >>> "fix" for a while longer:
>> >>>
>> >>> https://github.com/NixOS/nixpkgs/pull/12552
>> >>
>> >> This puill-request is huge, but for setuptools, it comes down that they
>> >> updated from 18.2 to 19.4.
>> >
>> > Sounds like we're going down the same road. I've started a branch with
>> > the earlier patch and a few other fixes. Is it ok to overwrite the
>> > existing 'python-updates' branch on Savannah?
>> 
>> There are some commits in python-updates that haven't made its way into
>> any other branches. I created a new branch 'python-tests' for this fix.
>> 
>> Please build and fix as much as possible. I'd like to get this merged
>> ASAP. @Leo can you start this branch on Hydra?
>
> Several of the commit messages were truncated because of lines that
> began with '#:tests?'. The '#' character creates a comment.

Oh my. Thanks for taking care of that, not sure how I missed it.

> I fixed the commit messages with `git rebase`. This means that most of
> the commits are now signed by me. I didn't change the commits
> themselves.
>
> The evaluation is pending:
>
> https://hydra.gnu.org/jobset/gnu/python-tests

Great, thanks! Looks like this is going to take a while. Once this is
done, should consider reviving 'python-updates' since some breakages
seem Python 3.5 specific.

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

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 1436 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Tue, Dec 13, 2016 at 09:21:44PM +0100, Marius Bakke wrote:
>> 
>> >>> Yikes, I had hoped to avoid addressing that Nix issue and the humongous
>> >>> "fix" for a while longer:
>> >>>
>> >>> https://github.com/NixOS/nixpkgs/pull/12552
>> >>
>> >> This puill-request is huge, but for setuptools, it comes down that they
>> >> updated from 18.2 to 19.4.
>> >
>> > Sounds like we're going down the same road. I've started a branch with
>> > the earlier patch and a few other fixes. Is it ok to overwrite the
>> > existing 'python-updates' branch on Savannah?
>> 
>> There are some commits in python-updates that haven't made its way into
>> any other branches. I created a new branch 'python-tests' for this fix.
>> 
>> Please build and fix as much as possible. I'd like to get this merged
>> ASAP. @Leo can you start this branch on Hydra?
>
> Several of the commit messages were truncated because of lines that
> began with '#:tests?'. The '#' character creates a comment.
>
> I fixed the commit messages with `git rebase`. This means that most of
> the commits are now signed by me. I didn't change the commits
> themselves.
>
> The evaluation is pending:
>
> https://hydra.gnu.org/jobset/gnu/python-tests

All failures from this run should be fixed, and then some. Can you
restart it?

@Hartmut, what was your command for building everything python? :-)

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 180 bytes --]

On Thu, Dec 15, 2016 at 10:26:45AM +0100, Marius Bakke wrote:
> All failures from this run should be fixed, and then some. Can you
> restart it?

Done, thanks for working on this!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Hartmut Goebel]

Am 15.12.2016 um 10:26 schrieb Marius Bakke:
> @Hartmut, what was your command for building everything python? :-)

I have a script for this :-) But maybe there is a more inteligent way like:

guix refresh -l python | sed 's/.*: //' | xargs guix build --keep-going

-- 
Regards
Hartmut Goebel

| Hartmut Goebel          | h.goebel@crazy-compilers.com               |
| www.crazy-compilers.com | compilers which you thought are impossible |

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 487 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Thu, Dec 15, 2016 at 10:26:45AM +0100, Marius Bakke wrote:
>> All failures from this run should be fixed, and then some. Can you
>> restart it?
>
> Done, thanks for working on this!

Most of the current failures should now be fixed as well. Do you think
another Hydra run is necessary? I would like to merge this before the
release, so I think any remaining failures can be handled directly on
master for more visibility and testing. WDYT?

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 1215 bytes --]

On Fri, Dec 16, 2016 at 03:02:20PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> 
> > On Thu, Dec 15, 2016 at 10:26:45AM +0100, Marius Bakke wrote:
> >> All failures from this run should be fixed, and then some. Can you
> >> restart it?
> >
> > Done, thanks for working on this!
> 
> Most of the current failures should now be fixed as well. Do you think
> another Hydra run is necessary? I would like to merge this before the
> release, so I think any remaining failures can be handled directly on
> master for more visibility and testing. WDYT?

There are still ~4000 queued builds:

https://hydra.gnu.org/eval/109396?full=1#tabs-unfinished

I'd like to wait for it to be closer to done before merging.

I think it's okay to release without these changes because, in practice,
the bug in question doesn't cause problems for people who are using our
packages, right? My understanding is that people installing from 0.12.0
won't have large numbers of Python packages that fail to build, since we
are dealing with spurious test successes.

It's certainly not desirable to release with this bug, but will there be
"real" problems for users?

What do you think?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 1788 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Fri, Dec 16, 2016 at 03:02:20PM +0100, Marius Bakke wrote:
>> Leo Famulari <leo@famulari.name> writes:
>> 
>> > On Thu, Dec 15, 2016 at 10:26:45AM +0100, Marius Bakke wrote:
>> >> All failures from this run should be fixed, and then some. Can you
>> >> restart it?
>> >
>> > Done, thanks for working on this!
>> 
>> Most of the current failures should now be fixed as well. Do you think
>> another Hydra run is necessary? I would like to merge this before the
>> release, so I think any remaining failures can be handled directly on
>> master for more visibility and testing. WDYT?
>
> There are still ~4000 queued builds:
>
> https://hydra.gnu.org/eval/109396?full=1#tabs-unfinished
>
> I'd like to wait for it to be closer to done before merging.

I think that job can be cancelled and a new one started, since there
have been quite a few fixes since. I've now successfully built
`./pre-inst-env guix build $(guix package -A python | awk '{print $1}')`
locally, but there may still be some leaf packages that are failing.

> I think it's okay to release without these changes because, in practice,
> the bug in question doesn't cause problems for people who are using our
> packages, right? My understanding is that people installing from 0.12.0
> won't have large numbers of Python packages that fail to build, since we
> are dealing with spurious test successes.
>
> It's certainly not desirable to release with this bug, but will there be
> "real" problems for users?
>
> What do you think?

Yes, it would have to be merged now or wait until after the release.
Very few actual problems have been solved, and none in core packages.
So let's postpone this and focus on getting master in shape :-)

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 965 bytes --]

On Sat, Dec 17, 2016 at 06:38:26PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> > https://hydra.gnu.org/eval/109396?full=1#tabs-unfinished
> >
> > I'd like to wait for it to be closer to done before merging.
> 
> I think that job can be cancelled and a new one started, since there
> have been quite a few fixes since. I've now successfully built
> `./pre-inst-env guix build $(guix package -A python | awk '{print $1}')`
> locally, but there may still be some leaf packages that are failing.

I've canceled that evaluation and started another.

> > It's certainly not desirable to release with this bug, but will there be
> > "real" problems for users?
> >
> > What do you think?
> 
> Yes, it would have to be merged now or wait until after the release.
> Very few actual problems have been solved, and none in core packages.
> So let's postpone this and focus on getting master in shape :-)

Okay, sounds good :)

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 758 bytes --]

On Sat, Dec 17, 2016 at 12:56:21PM -0500, Leo Famulari wrote:
> On Sat, Dec 17, 2016 at 06:38:26PM +0100, Marius Bakke wrote:
> > Leo Famulari <leo@famulari.name> writes:
> > > https://hydra.gnu.org/eval/109396?full=1#tabs-unfinished
> > >
> > > I'd like to wait for it to be closer to done before merging.
> > 
> > I think that job can be cancelled and a new one started, since there
> > have been quite a few fixes since. I've now successfully built
> > `./pre-inst-env guix build $(guix package -A python | awk '{print $1}')`
> > locally, but there may still be some leaf packages that are failing.
> 
> I've canceled that evaluation and started another.

The evaluation failed:

https://hydra.gnu.org/jobset/gnu/python-tests#tabs-errors

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 884 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Sat, Dec 17, 2016 at 12:56:21PM -0500, Leo Famulari wrote:
>> On Sat, Dec 17, 2016 at 06:38:26PM +0100, Marius Bakke wrote:
>> > Leo Famulari <leo@famulari.name> writes:
>> > > https://hydra.gnu.org/eval/109396?full=1#tabs-unfinished
>> > >
>> > > I'd like to wait for it to be closer to done before merging.
>> > 
>> > I think that job can be cancelled and a new one started, since there
>> > have been quite a few fixes since. I've now successfully built
>> > `./pre-inst-env guix build $(guix package -A python | awk '{print $1}')`
>> > locally, but there may still be some leaf packages that are failing.
>> 
>> I've canceled that evaluation and started another.
>
> The evaluation failed:
>
> https://hydra.gnu.org/jobset/gnu/python-tests#tabs-errors

Whoops. Fixed in 7b6ff42aa3cf9398213d3c1f0e5f87c45a9730df.

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 160 bytes --]

On Sun, Dec 18, 2016 at 12:47:59AM +0100, Marius Bakke wrote:
> Whoops. Fixed in 7b6ff42aa3cf9398213d3c1f0e5f87c45a9730df.

Thanks! I started a new evaluation.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 299 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Sun, Dec 18, 2016 at 12:47:59AM +0100, Marius Bakke wrote:
>> Whoops. Fixed in 7b6ff42aa3cf9398213d3c1f0e5f87c45a9730df.
>
> Thanks! I started a new evaluation.

Looks like this one timed out:

https://hydra.gnu.org/jobset/gnu/python-tests#tabs-errors

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 421 bytes --]

On Sun, Dec 18, 2016 at 07:55:54PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> 
> > On Sun, Dec 18, 2016 at 12:47:59AM +0100, Marius Bakke wrote:
> >> Whoops. Fixed in 7b6ff42aa3cf9398213d3c1f0e5f87c45a9730df.
> >
> > Thanks! I started a new evaluation.
> 
> Looks like this one timed out:
> 
> https://hydra.gnu.org/jobset/gnu/python-tests#tabs-errors

Bah, I started a new one.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 694 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Sun, Dec 18, 2016 at 07:55:54PM +0100, Marius Bakke wrote:
>> Leo Famulari <leo@famulari.name> writes:
>> 
>> > On Sun, Dec 18, 2016 at 12:47:59AM +0100, Marius Bakke wrote:
>> >> Whoops. Fixed in 7b6ff42aa3cf9398213d3c1f0e5f87c45a9730df.
>> >
>> > Thanks! I started a new evaluation.
>> 
>> Looks like this one timed out:
>> 
>> https://hydra.gnu.org/jobset/gnu/python-tests#tabs-errors
>
> Bah, I started a new one.

Did this one get cancelled? Perhaps we should merge this to 'staging'
and deal with any remaining failures when that is ready to be evaluated.

I believe the vast majority of python packages are now "fixed".

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 670 bytes --]

On Tue, Dec 20, 2016 at 07:49:59PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> > Bah, I started a new one.
> 
> Did this one get cancelled? Perhaps we should merge this to 'staging'
> and deal with any remaining failures when that is ready to be evaluated.

Yes, it looks like that, although I don't know the exact reason. Hydra
has lots of resource management and stability problems.

I guess the python-tests jobset was canceled to free some resources for
the release-0.12.0 jobset.

I'd like to restart python-tests after 0.12.0 is released.

> I believe the vast majority of python packages are now "fixed".

Thank you!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 708 bytes --]

Leo Famulari <leo@famulari.name> writes:

> On Tue, Dec 20, 2016 at 07:49:59PM +0100, Marius Bakke wrote:
>> Leo Famulari <leo@famulari.name> writes:
>> > Bah, I started a new one.
>> 
>> Did this one get cancelled? Perhaps we should merge this to 'staging'
>> and deal with any remaining failures when that is ready to be evaluated.
>
> Yes, it looks like that, although I don't know the exact reason. Hydra
> has lots of resource management and stability problems.
>
> I guess the python-tests jobset was canceled to free some resources for
> the release-0.12.0 jobset.
>
> I'd like to restart python-tests after 0.12.0 is released.

Should we try to evaluate this branch before staging?

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

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 848 bytes --]

On Sat, Dec 24, 2016 at 02:43:38PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> 
> > On Tue, Dec 20, 2016 at 07:49:59PM +0100, Marius Bakke wrote:
> >> Leo Famulari <leo@famulari.name> writes:
> >> > Bah, I started a new one.
> >> 
> >> Did this one get cancelled? Perhaps we should merge this to 'staging'
> >> and deal with any remaining failures when that is ready to be evaluated.
> >
> > Yes, it looks like that, although I don't know the exact reason. Hydra
> > has lots of resource management and stability problems.
> >
> > I guess the python-tests jobset was canceled to free some resources for
> > the release-0.12.0 jobset.
> >
> > I'd like to restart python-tests after 0.12.0 is released.
> 
> Should we try to evaluate this branch before staging?

Yes, I just started a new evaluation.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#25177: Test failures don't cause some Python packages to fail

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 437 bytes --]

On Sat, Dec 24, 2016 at 11:26:37AM -0500, Leo Famulari wrote:
> On Sat, Dec 24, 2016 at 02:43:38PM +0100, Marius Bakke wrote:
> > Should we try to evaluate this branch before staging?
> 
> Yes, I just started a new evaluation.

There are still some failing packages that we need to address:

https://hydra.gnu.org/eval/109407?compare=master&full=1

... but I haven't had time to look at them individually.

In a few days :)

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]