[Matplotlib-devel] v1.5.0rc2

classic Classic list List threaded Threaded
17 messages Options
Reply | Threaded
Open this post in threaded view
|

[Matplotlib-devel] v1.5.0rc2

tcaswell
Hey folks,

I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully the new conda packages will be available on `-c conda-forge` within the next day or so.

Please open any new PRs targetted for 1.5.0 against the v1.5.x branch and remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when you merge them (or ping me to do it).  

Feel free to begin merging new features into master again.

In addition to the normal release notes we should plan to also write several blog posts for the numfocus blog.  I think it would be good to have a post for at least each of 
  - property cycling
  - labeled data plotting
  - auto-redraw
  - %matplotlib notebook
  - the new color maps
  - the style module + new style sheets

Any volunteers to write those?  

Tom

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Matthew Brett
Hi,

On Fri, Oct 2, 2015 at 4:04 PM, Thomas Caswell <[hidden email]> wrote:
> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully
> the new conda packages will be available on `-c conda-forge` within the next
> day or so.

Wheels at http://wheels.scipy.org/

pip install --pre -f http://wheels.scipy.org matplotlib

Matthew
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
Thanks Matthew!

Tom

On Fri, Oct 2, 2015 at 7:39 PM Matthew Brett <[hidden email]> wrote:
Hi,

On Fri, Oct 2, 2015 at 4:04 PM, Thomas Caswell <[hidden email]> wrote:
> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully
> the new conda packages will be available on `-c conda-forge` within the next
> day or so.

Wheels at http://wheels.scipy.org/

pip install --pre -f http://wheels.scipy.org matplotlib

Matthew

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
In reply to this post by tcaswell
Hey! I'm working to get this into Debian, so I'd like to understand
how strong the dependency against cycler is: is there a way to opt-out
from it at the moment? We dont have it already available, and avoiding
that dependency will help us the mpl in Debian (which in turn will
help Debian migrate to python3.5)

Thanks!

On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]> wrote:

> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully
> the new conda packages will be available on `-c conda-forge` within the next
> day or so.
>
> Please open any new PRs targetted for 1.5.0 against the v1.5.x branch and
> remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when you
> merge them (or ping me to do it).
>
> Feel free to begin merging new features into master again.
>
> In addition to the normal release notes we should plan to also write several
> blog posts for the numfocus blog.  I think it would be good to have a post
> for at least each of
>   - property cycling
>   - labeled data plotting
>   - auto-redraw
>   - %matplotlib notebook
>   - the new color maps
>   - the style module + new style sheets
>
> Any volunteers to write those?
>
> Tom
>
> _______________________________________________
> Matplotlib-devel mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-devel
>



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
It is a hard dependency, but it is only a few hundred line python file.  

If getting it in as a top-level debian package is not possible, it should not be too hard to vendor it.

My preference would be if you could handle this as a debian-specific vendoring, but if this is really a blocker for debian (at it's down-stream ecosystem) we can work something out up upstream.

When we started working on cycler part of the discussion was to split it off because down-stream packagers liked to split things up!

Tom

On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
Hey! I'm working to get this into Debian, so I'd like to understand
how strong the dependency against cycler is: is there a way to opt-out
from it at the moment? We dont have it already available, and avoiding
that dependency will help us the mpl in Debian (which in turn will
help Debian migrate to python3.5)

Thanks!

On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]> wrote:
> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully
> the new conda packages will be available on `-c conda-forge` within the next
> day or so.
>
> Please open any new PRs targetted for 1.5.0 against the v1.5.x branch and
> remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when you
> merge them (or ping me to do it).
>
> Feel free to begin merging new features into master again.
>
> In addition to the normal release notes we should plan to also write several
> blog posts for the numfocus blog.  I think it would be good to have a post
> for at least each of
>   - property cycling
>   - labeled data plotting
>   - auto-redraw
>   - %matplotlib notebook
>   - the new color maps
>   - the style module + new style sheets
>
> Any volunteers to write those?
>
> Tom
>
> _______________________________________________
> Matplotlib-devel mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-devel
>



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Christoph Gohlke
In reply to this post by tcaswell
On 10/2/2015 4:04 PM, Thomas Caswell wrote:

> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.
> Hopefully the new conda packages will be available on `-c conda-forge`
> within the next day or so.
>
> Please open any new PRs targetted for 1.5.0 against the v1.5.x branch
> and remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x
> when you merge them (or ping me to do it).
>
> Feel free to begin merging new features into master again.
>
> In addition to the normal release notes we should plan to also write
> several blog posts for the numfocus blog.  I think it would be good to
> have a post for at least each of
>    - property cycling
>    - labeled data plotting
>    - auto-redraw
>    - %matplotlib notebook
>    - the new color maps
>    - the style module + new style sheets
>
> Any volunteers to write those?
>
> Tom
>
>

Forgot to mention: Windows installers, wheels and the HTML help file are
at
<http://sourceforge.net/projects/matplotlib/files/matplotlib/matplotlib-1.5.0rc2/windows/>

Christoph
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

OceanWolf
In reply to this post by tcaswell
As a debian user myself I would not have thought it impossible, it just takes longer, new packages can take several months to enter into unstable, but once in the process goes relatively fast.

I speak here only as an observer of the process, I of course defer to Sandro for the details (p.s. congratulations for holding and getting 1.4.x into jessie at the 11th hour).

On 04/10/15 04:43, Thomas Caswell wrote:
It is a hard dependency, but it is only a few hundred line python file.  

If getting it in as a top-level debian package is not possible, it should not be too hard to vendor it.

My preference would be if you could handle this as a debian-specific vendoring, but if this is really a blocker for debian (at it's down-stream ecosystem) we can work something out up upstream.

When we started working on cycler part of the discussion was to split it off because down-stream packagers liked to split things up!

Tom

On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
Hey! I'm working to get this into Debian, so I'd like to understand
how strong the dependency against cycler is: is there a way to opt-out
from it at the moment? We dont have it already available, and avoiding
that dependency will help us the mpl in Debian (which in turn will
help Debian migrate to python3.5)

Thanks!

On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]> wrote:
> Hey folks,
>
> I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.  Hopefully
> the new conda packages will be available on `-c conda-forge` within the next
> day or so.
>
> Please open any new PRs targetted for 1.5.0 against the v1.5.x branch and
> remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when you
> merge them (or ping me to do it).
>
> Feel free to begin merging new features into master again.
>
> In addition to the normal release notes we should plan to also write several
> blog posts for the numfocus blog.  I think it would be good to have a post
> for at least each of
>   - property cycling
>   - labeled data plotting
>   - auto-redraw
>   - %matplotlib notebook
>   - the new color maps
>   - the style module + new style sheets
>
> Any volunteers to write those?
>
> Tom
>
> _______________________________________________
> Matplotlib-devel mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-devel
>



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi


_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel


_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
In reply to this post by tcaswell
On Sun, Oct 4, 2015 at 3:43 AM, Thomas Caswell <[hidden email]> wrote:
> It is a hard dependency, but it is only a few hundred line python file.

that's ok, I'll get it packaged and will ask to be fast tracked for acceptance.

> If getting it in as a top-level debian package is not possible, it should
> not be too hard to vendor it.
>
> My preference would be if you could handle this as a debian-specific
> vendoring, but if this is really a blocker for debian (at it's down-stream
> ecosystem) we can work something out up upstream.
>
> When we started working on cycler part of the discussion was to split it off
> because down-stream packagers liked to split things up!

hehe, yeah we usually do (even though it sometimes makes our own lives
harder ;) )

thanks for the quick reply!

>
> Tom
>
> On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
>>
>> Hey! I'm working to get this into Debian, so I'd like to understand
>> how strong the dependency against cycler is: is there a way to opt-out
>> from it at the moment? We dont have it already available, and avoiding
>> that dependency will help us the mpl in Debian (which in turn will
>> help Debian migrate to python3.5)
>>
>> Thanks!
>>
>> On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]>
>> wrote:
>> > Hey folks,
>> >
>> > I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.
>> > Hopefully
>> > the new conda packages will be available on `-c conda-forge` within the
>> > next
>> > day or so.
>> >
>> > Please open any new PRs targetted for 1.5.0 against the v1.5.x branch
>> > and
>> > remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when
>> > you
>> > merge them (or ping me to do it).
>> >
>> > Feel free to begin merging new features into master again.
>> >
>> > In addition to the normal release notes we should plan to also write
>> > several
>> > blog posts for the numfocus blog.  I think it would be good to have a
>> > post
>> > for at least each of
>> >   - property cycling
>> >   - labeled data plotting
>> >   - auto-redraw
>> >   - %matplotlib notebook
>> >   - the new color maps
>> >   - the style module + new style sheets
>> >
>> > Any volunteers to write those?
>> >
>> > Tom
>> >
>> > _______________________________________________
>> > Matplotlib-devel mailing list
>> > [hidden email]
>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>> >
>>
>>
>>
>> --
>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>> My website: http://matrixhasu.altervista.org/
>> Me at Debian: http://wiki.debian.org/SandroTosi



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
Ok, i managed to reach the point where I can build mpl on debian, but
I'm getting some weird unittest failures (cut&paste will suck, so I'm
also attaching the full build log):

```
ERROR: matplotlib.tests.test_style.test_use_url
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_style.py",
line 61, in test_use_url
    with style.context('https://gist.github.com/adrn/6590261/raw'):
  File "/usr/lib/python2.7/contextlib.py", line 17, in __enter__
    return self.gen.next()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 121, in context
    use(style)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 90, in use
    raise IOError(msg % style)
IOError: 'https://gist.github.com/adrn/6590261/raw' not found in the
style library and input is not a valid URL or path. See
`style.available` for list of available styles.
```

it seems the test is trying to reach a remote URL (and fails): in
debian we dont allow package to reach the internet when building, any
suggestions how to disable this test?

```
ERROR: Failure: RuntimeError (PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 409, in
loadTestsFromName
    module = resolve_name(addr.module)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 312, in
resolve_name
    module = __import__('.'.join(parts_copy))
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_backend_qt4.py",
line 19, in <module>
    from matplotlib.backends.qt_compat import QtCore
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/backends/qt_compat.py",
line 144, in <module>
    raise RuntimeError("PyQt{4,5} bindings found despite sip importing\n"
RuntimeError: PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.
```

I'm not sure I understand what it's trying to say. there are indeed by
Qt4 and Qt5 available as long as Sip (as it's a dependency of PyQt5),
so what is the problem there? note it worked with 1.4.x with these
packges set.

```
ERROR: test suite for <class
'matplotlib.sphinxext.tests.test_tinypages.TestTinyPages'>
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 209, in run
    self.setUp()
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 292, in setUp
    self.setupContext(ancestor)
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 315, in
setupContext
    try_run(context, names)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 471, in try_run
    return func()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7-pydebug/matplotlib/sphinxext/tests/test_tinypages.py",
line 58, in setup_class
    out, err))
RuntimeError: sphinx-build failed with stdout:
Running Sphinx v1.3.1
making output directory...

stderr:

Extension error:
Could not import extension matplotlib.sphinxext.plot_directive
(exception: No module named _path)
```

I dont know what this _path should be imported  from, as after
grepping a bit I was not able to identify that.

let me know if you want me to report them as bugs or this email is fine.


On Sun, Oct 4, 2015 at 12:47 PM, Sandro Tosi <[hidden email]> wrote:

> On Sun, Oct 4, 2015 at 3:43 AM, Thomas Caswell <[hidden email]> wrote:
>> It is a hard dependency, but it is only a few hundred line python file.
>
> that's ok, I'll get it packaged and will ask to be fast tracked for acceptance.
>
>> If getting it in as a top-level debian package is not possible, it should
>> not be too hard to vendor it.
>>
>> My preference would be if you could handle this as a debian-specific
>> vendoring, but if this is really a blocker for debian (at it's down-stream
>> ecosystem) we can work something out up upstream.
>>
>> When we started working on cycler part of the discussion was to split it off
>> because down-stream packagers liked to split things up!
>
> hehe, yeah we usually do (even though it sometimes makes our own lives
> harder ;) )
>
> thanks for the quick reply!
>
>>
>> Tom
>>
>> On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
>>>
>>> Hey! I'm working to get this into Debian, so I'd like to understand
>>> how strong the dependency against cycler is: is there a way to opt-out
>>> from it at the moment? We dont have it already available, and avoiding
>>> that dependency will help us the mpl in Debian (which in turn will
>>> help Debian migrate to python3.5)
>>>
>>> Thanks!
>>>
>>> On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]>
>>> wrote:
>>> > Hey folks,
>>> >
>>> > I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.
>>> > Hopefully
>>> > the new conda packages will be available on `-c conda-forge` within the
>>> > next
>>> > day or so.
>>> >
>>> > Please open any new PRs targetted for 1.5.0 against the v1.5.x branch
>>> > and
>>> > remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when
>>> > you
>>> > merge them (or ping me to do it).
>>> >
>>> > Feel free to begin merging new features into master again.
>>> >
>>> > In addition to the normal release notes we should plan to also write
>>> > several
>>> > blog posts for the numfocus blog.  I think it would be good to have a
>>> > post
>>> > for at least each of
>>> >   - property cycling
>>> >   - labeled data plotting
>>> >   - auto-redraw
>>> >   - %matplotlib notebook
>>> >   - the new color maps
>>> >   - the style module + new style sheets
>>> >
>>> > Any volunteers to write those?
>>> >
>>> > Tom
>>> >
>>> > _______________________________________________
>>> > Matplotlib-devel mailing list
>>> > [hidden email]
>>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>>> >
>>>
>>>
>>>
>>> --
>>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>>> My website: http://matrixhasu.altervista.org/
>>> Me at Debian: http://wiki.debian.org/SandroTosi
>
>
>
> --
> Sandro Tosi (aka morph, morpheus, matrixhasu)
> My website: http://matrixhasu.altervista.org/
> Me at Debian: http://wiki.debian.org/SandroTosi


--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel

matplotlib_1.5.0~rc2-1_amd64.build.bz2 (116K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
The first one is that one of our tests reaches out to the internet for a test.  We had a rough idea that this was not acceptable for debian, but were not sure.  That test probably just needs to be wrapped in a know fail if the internet is not accessible.

The second one is a bug that I put in just before the RC2, sorry.

`_path` is one of the c-extensions we build, there should be a `_path.so.*` in the top-level mpl folder.

Tom

On Sun, Oct 4, 2015 at 3:43 PM Sandro Tosi <[hidden email]> wrote:
Ok, i managed to reach the point where I can build mpl on debian, but
I'm getting some weird unittest failures (cut&paste will suck, so I'm
also attaching the full build log):

```
ERROR: matplotlib.tests.test_style.test_use_url
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_style.py",
line 61, in test_use_url
    with style.context('https://gist.github.com/adrn/6590261/raw'):
  File "/usr/lib/python2.7/contextlib.py", line 17, in __enter__
    return self.gen.next()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 121, in context
    use(style)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 90, in use
    raise IOError(msg % style)
IOError: 'https://gist.github.com/adrn/6590261/raw' not found in the
style library and input is not a valid URL or path. See
`style.available` for list of available styles.
```

it seems the test is trying to reach a remote URL (and fails): in
debian we dont allow package to reach the internet when building, any
suggestions how to disable this test?

```
ERROR: Failure: RuntimeError (PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 409, in
loadTestsFromName
    module = resolve_name(addr.module)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 312, in
resolve_name
    module = __import__('.'.join(parts_copy))
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_backend_qt4.py",
line 19, in <module>
    from matplotlib.backends.qt_compat import QtCore
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/backends/qt_compat.py",
line 144, in <module>
    raise RuntimeError("PyQt{4,5} bindings found despite sip importing\n"
RuntimeError: PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.
```

I'm not sure I understand what it's trying to say. there are indeed by
Qt4 and Qt5 available as long as Sip (as it's a dependency of PyQt5),
so what is the problem there? note it worked with 1.4.x with these
packges set.

```
ERROR: test suite for <class
'matplotlib.sphinxext.tests.test_tinypages.TestTinyPages'>
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 209, in run
    self.setUp()
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 292, in setUp
    self.setupContext(ancestor)
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 315, in
setupContext
    try_run(context, names)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 471, in try_run
    return func()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7-pydebug/matplotlib/sphinxext/tests/test_tinypages.py",
line 58, in setup_class
    out, err))
RuntimeError: sphinx-build failed with stdout:
Running Sphinx v1.3.1
making output directory...

stderr:

Extension error:
Could not import extension matplotlib.sphinxext.plot_directive
(exception: No module named _path)
```

I dont know what this _path should be imported  from, as after
grepping a bit I was not able to identify that.

let me know if you want me to report them as bugs or this email is fine.


On Sun, Oct 4, 2015 at 12:47 PM, Sandro Tosi <[hidden email]> wrote:
> On Sun, Oct 4, 2015 at 3:43 AM, Thomas Caswell <[hidden email]> wrote:
>> It is a hard dependency, but it is only a few hundred line python file.
>
> that's ok, I'll get it packaged and will ask to be fast tracked for acceptance.
>
>> If getting it in as a top-level debian package is not possible, it should
>> not be too hard to vendor it.
>>
>> My preference would be if you could handle this as a debian-specific
>> vendoring, but if this is really a blocker for debian (at it's down-stream
>> ecosystem) we can work something out up upstream.
>>
>> When we started working on cycler part of the discussion was to split it off
>> because down-stream packagers liked to split things up!
>
> hehe, yeah we usually do (even though it sometimes makes our own lives
> harder ;) )
>
> thanks for the quick reply!
>
>>
>> Tom
>>
>> On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
>>>
>>> Hey! I'm working to get this into Debian, so I'd like to understand
>>> how strong the dependency against cycler is: is there a way to opt-out
>>> from it at the moment? We dont have it already available, and avoiding
>>> that dependency will help us the mpl in Debian (which in turn will
>>> help Debian migrate to python3.5)
>>>
>>> Thanks!
>>>
>>> On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]>
>>> wrote:
>>> > Hey folks,
>>> >
>>> > I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.
>>> > Hopefully
>>> > the new conda packages will be available on `-c conda-forge` within the
>>> > next
>>> > day or so.
>>> >
>>> > Please open any new PRs targetted for 1.5.0 against the v1.5.x branch
>>> > and
>>> > remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when
>>> > you
>>> > merge them (or ping me to do it).
>>> >
>>> > Feel free to begin merging new features into master again.
>>> >
>>> > In addition to the normal release notes we should plan to also write
>>> > several
>>> > blog posts for the numfocus blog.  I think it would be good to have a
>>> > post
>>> > for at least each of
>>> >   - property cycling
>>> >   - labeled data plotting
>>> >   - auto-redraw
>>> >   - %matplotlib notebook
>>> >   - the new color maps
>>> >   - the style module + new style sheets
>>> >
>>> > Any volunteers to write those?
>>> >
>>> > Tom
>>> >
>>> > _______________________________________________
>>> > Matplotlib-devel mailing list
>>> > [hidden email]
>>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>>> >
>>>
>>>
>>>
>>> --
>>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>>> My website: http://matrixhasu.altervista.org/
>>> Me at Debian: http://wiki.debian.org/SandroTosi
>
>
>
> --
> Sandro Tosi (aka morph, morpheus, matrixhasu)
> My website: http://matrixhasu.altervista.org/
> Me at Debian: http://wiki.debian.org/SandroTosi



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
This PR should address the url issue: https://github.com/matplotlib/matplotlib/pull/5188
This PR should address the qt issue: https://github.com/matplotlib/matplotlib/pull/5187



On Sun, Oct 4, 2015 at 3:51 PM Thomas Caswell <[hidden email]> wrote:
The first one is that one of our tests reaches out to the internet for a test.  We had a rough idea that this was not acceptable for debian, but were not sure.  That test probably just needs to be wrapped in a know fail if the internet is not accessible.

The second one is a bug that I put in just before the RC2, sorry.

`_path` is one of the c-extensions we build, there should be a `_path.so.*` in the top-level mpl folder.

Tom

On Sun, Oct 4, 2015 at 3:43 PM Sandro Tosi <[hidden email]> wrote:
Ok, i managed to reach the point where I can build mpl on debian, but
I'm getting some weird unittest failures (cut&paste will suck, so I'm
also attaching the full build log):

```
ERROR: matplotlib.tests.test_style.test_use_url
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_style.py",
line 61, in test_use_url
    with style.context('https://gist.github.com/adrn/6590261/raw'):
  File "/usr/lib/python2.7/contextlib.py", line 17, in __enter__
    return self.gen.next()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 121, in context
    use(style)
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/style/core.py",
line 90, in use
    raise IOError(msg % style)
IOError: 'https://gist.github.com/adrn/6590261/raw' not found in the
style library and input is not a valid URL or path. See
`style.available` for list of available styles.
```

it seems the test is trying to reach a remote URL (and fails): in
debian we dont allow package to reach the internet when building, any
suggestions how to disable this test?

```
ERROR: Failure: RuntimeError (PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 409, in
loadTestsFromName
    module = resolve_name(addr.module)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 312, in
resolve_name
    module = __import__('.'.join(parts_copy))
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/tests/test_backend_qt4.py",
line 19, in <module>
    from matplotlib.backends.qt_compat import QtCore
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7/matplotlib/backends/qt_compat.py",
line 144, in <module>
    raise RuntimeError("PyQt{4,5} bindings found despite sip importing\n"
RuntimeError: PyQt{4,5} bindings found despite sip importing
Please install PyQt4 or PyQt5, uninstall sip or explicitly set the
pyside backend.
```

I'm not sure I understand what it's trying to say. there are indeed by
Qt4 and Qt5 available as long as Sip (as it's a dependency of PyQt5),
so what is the problem there? note it worked with 1.4.x with these
packges set.

```
ERROR: test suite for <class
'matplotlib.sphinxext.tests.test_tinypages.TestTinyPages'>
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 209, in run
    self.setUp()
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 292, in setUp
    self.setupContext(ancestor)
  File "/usr/lib/python2.7/dist-packages/nose/suite.py", line 315, in
setupContext
    try_run(context, names)
  File "/usr/lib/python2.7/dist-packages/nose/util.py", line 471, in try_run
    return func()
  File "/tmp/buildd/matplotlib-1.5.0~rc2/build/lib.linux-x86_64-2.7-pydebug/matplotlib/sphinxext/tests/test_tinypages.py",
line 58, in setup_class
    out, err))
RuntimeError: sphinx-build failed with stdout:
Running Sphinx v1.3.1
making output directory...

stderr:

Extension error:
Could not import extension matplotlib.sphinxext.plot_directive
(exception: No module named _path)
```

I dont know what this _path should be imported  from, as after
grepping a bit I was not able to identify that.

let me know if you want me to report them as bugs or this email is fine.


On Sun, Oct 4, 2015 at 12:47 PM, Sandro Tosi <[hidden email]> wrote:
> On Sun, Oct 4, 2015 at 3:43 AM, Thomas Caswell <[hidden email]> wrote:
>> It is a hard dependency, but it is only a few hundred line python file.
>
> that's ok, I'll get it packaged and will ask to be fast tracked for acceptance.
>
>> If getting it in as a top-level debian package is not possible, it should
>> not be too hard to vendor it.
>>
>> My preference would be if you could handle this as a debian-specific
>> vendoring, but if this is really a blocker for debian (at it's down-stream
>> ecosystem) we can work something out up upstream.
>>
>> When we started working on cycler part of the discussion was to split it off
>> because down-stream packagers liked to split things up!
>
> hehe, yeah we usually do (even though it sometimes makes our own lives
> harder ;) )
>
> thanks for the quick reply!
>
>>
>> Tom
>>
>> On Sat, Oct 3, 2015 at 8:02 PM Sandro Tosi <[hidden email]> wrote:
>>>
>>> Hey! I'm working to get this into Debian, so I'd like to understand
>>> how strong the dependency against cycler is: is there a way to opt-out
>>> from it at the moment? We dont have it already available, and avoiding
>>> that dependency will help us the mpl in Debian (which in turn will
>>> help Debian migrate to python3.5)
>>>
>>> Thanks!
>>>
>>> On Sat, Oct 3, 2015 at 12:04 AM, Thomas Caswell <[hidden email]>
>>> wrote:
>>> > Hey folks,
>>> >
>>> > I have tagged v1.5.0rc2 and created a v1.5.x branch on upstream.
>>> > Hopefully
>>> > the new conda packages will be available on `-c conda-forge` within the
>>> > next
>>> > day or so.
>>> >
>>> > Please open any new PRs targetted for 1.5.0 against the v1.5.x branch
>>> > and
>>> > remember to cherry pick the few open PRs for 1.5.0 back to 1.5.x when
>>> > you
>>> > merge them (or ping me to do it).
>>> >
>>> > Feel free to begin merging new features into master again.
>>> >
>>> > In addition to the normal release notes we should plan to also write
>>> > several
>>> > blog posts for the numfocus blog.  I think it would be good to have a
>>> > post
>>> > for at least each of
>>> >   - property cycling
>>> >   - labeled data plotting
>>> >   - auto-redraw
>>> >   - %matplotlib notebook
>>> >   - the new color maps
>>> >   - the style module + new style sheets
>>> >
>>> > Any volunteers to write those?
>>> >
>>> > Tom
>>> >
>>> > _______________________________________________
>>> > Matplotlib-devel mailing list
>>> > [hidden email]
>>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>>> >
>>>
>>>
>>>
>>> --
>>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>>> My website: http://matrixhasu.altervista.org/
>>> Me at Debian: http://wiki.debian.org/SandroTosi
>
>
>
> --
> Sandro Tosi (aka morph, morpheus, matrixhasu)
> My website: http://matrixhasu.altervista.org/
> Me at Debian: http://wiki.debian.org/SandroTosi



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
On Sun, Oct 4, 2015 at 9:27 PM, Thomas Caswell <[hidden email]> wrote:
> On Sun, Oct 4, 2015 at 3:51 PM Thomas Caswell <[hidden email]> wrote:
>>
>> The first one is that one of our tests reaches out to the internet for a
>> test.  We had a rough idea that this was not acceptable for debian, but were
>> not sure.  That test probably just needs to be wrapped in a know fail if the
>> internet is not accessible.

yeah I seem to remember there was a msg asking for my input on it and
I falied to reply, sorry about that

> This PR should address the url issue:
> https://github.com/matplotlib/matplotlib/pull/5188

geez you're too quick! what I wanted to say is that we dont want to
anything contacting the internet, not that we actively prevent it. The
error in this case is probably that HTTPS is not correctly recognized
as a URL as you might need to use an SSL handler.

what I have in mind is a way to mark those tests as accessing the
internet and skipping them when running the unittest (something like
I'm doing at http://anonscm.debian.org/cgit/reportbug/reportbug.git/tree/test/test_urlutils.py#n10
and http://anonscm.debian.org/cgit/reportbug/reportbug.git/tree/Makefile#n12
but I am using the nosetest runner there).

this is how I run tests for mpl in debian:

ln 39-56 of http://anonscm.debian.org/viewvc/python-modules/packages/matplotlib/trunk/debian/rules?revision=34251&view=markup

>> The second one is a bug that I put in just before the RC2, sorry.

> This PR should address the qt issue:
> https://github.com/matplotlib/matplotlib/pull/5187

will that this in a bit

>> `_path` is one of the c-extensions we build, there should be a
>> `_path.so.*` in the top-level mpl folder.

I see the python extensions build in matplotlib/_path.so but then it
is not installed right? then that's the problem, as I run setup.py
install in a temp dir and then set the PYTHONPATH to that dir to pick
up the module.

in the meantime a new error came up, while building latex doc:

```
Adding blank page after the table of contents.
[1] [2] [1] [2]
Chapter 1.
(/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)

! Package inputenc Error: Unicode char \u8:́ not set up for use with LaTeX.

See the inputenc package documentation for explanation.
Type  H <return>  for immediate help.
 ...

l.163 Sté
           fan van der Walt's talk from SciPy2015
?
! Emergency stop.
 ...

l.163 Sté
           fan van der Walt's talk from SciPy2015
!  ==> Fatal error occurred, no output PDF file produced!
```

attached the Matplotlib.log file of the latex compiler

Thanks!

--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel

Matplotlib.log.bz2 (12K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Jouni K. Seppänen
Sandro Tosi <[hidden email]> writes:

> in the meantime a new error came up, while building latex doc:
>
> ```
> Adding blank page after the table of contents.
> [1] [2] [1] [2]
> Chapter 1.
> (/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)
>
> ! Package inputenc Error: Unicode char \u8:́ not set up for use with LaTeX.
>
> See the inputenc package documentation for explanation.
> Type  H <return>  for immediate help.
>  ...
>
> l.163 Sté
>            fan van der Walt's talk from SciPy2015

The letter é in the source file is encoded using U+0301 COMBINING ACUTE
ACCENT (UTF-8 bytes cc 81):

$ git grep 'van der W' | hexdump -C
00000000  73 74 79 6c 65 5f 63 68  61 6e 67 65 73 2e 72 73  |style_changes.rs|
00000010  74 3a 53 74 65 cc 81 66  61 6e 20 76 61 6e 20 64  |t:Ste..fan van d|
00000020  65 72 20 57 61 6c 74 27  73 20 74 61 6c 6b 20 66  |er Walt's talk f|
00000030  72 6f 6d 20 53 63 69 50  79 32 30 31 35 0a        |rom SciPy2015.|
0000003e

LaTeX would probably understand U+00E9 LATIN SMALL LETTER E WITH ACUTE
better. I'll make a PR.

--
Jouni K. Seppänen
http://www.iki.fi/jks

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
first of all, thanks a lot for the prompt support in making mpl ready
to be uploaded to Debian! which happened last night, and it is mostly
built on all our official architectures (except mips, it's taking its
time to be built there eheh), but there are some unittest failures,
how do you want me to report them: one issue per arcihtecture? all of
them together? something else? I expect most of them will be wontfix,
but some might uncover hidden bugs

On Mon, Oct 5, 2015 at 5:32 AM, Jouni K. Seppänen <[hidden email]> wrote:

> Sandro Tosi <[hidden email]> writes:
>
>> in the meantime a new error came up, while building latex doc:
>>
>> ```
>> Adding blank page after the table of contents.
>> [1] [2] [1] [2]
>> Chapter 1.
>> (/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)
>>
>> ! Package inputenc Error: Unicode char \u8:́ not set up for use with LaTeX.
>>
>> See the inputenc package documentation for explanation.
>> Type  H <return>  for immediate help.
>>  ...
>>
>> l.163 Sté
>>            fan van der Walt's talk from SciPy2015
>
> The letter é in the source file is encoded using U+0301 COMBINING ACUTE
> ACCENT (UTF-8 bytes cc 81):
>
> $ git grep 'van der W' | hexdump -C
> 00000000  73 74 79 6c 65 5f 63 68  61 6e 67 65 73 2e 72 73  |style_changes.rs|
> 00000010  74 3a 53 74 65 cc 81 66  61 6e 20 76 61 6e 20 64  |t:Ste..fan van d|
> 00000020  65 72 20 57 61 6c 74 27  73 20 74 61 6c 6b 20 66  |er Walt's talk f|
> 00000030  72 6f 6d 20 53 63 69 50  79 32 30 31 35 0a        |rom SciPy2015.|
> 0000003e
>
> LaTeX would probably understand U+00E9 LATIN SMALL LETTER E WITH ACUTE
> better. I'll make a PR.
>
> --
> Jouni K. Seppänen
> http://www.iki.fi/jks
>
> _______________________________________________
> Matplotlib-devel mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-devel

--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
Can you point me where the logs are?  

I suspect one issue per arch is the best approach.

Tom

On Tue, Oct 6, 2015 at 6:57 PM Sandro Tosi <[hidden email]> wrote:
first of all, thanks a lot for the prompt support in making mpl ready
to be uploaded to Debian! which happened last night, and it is mostly
built on all our official architectures (except mips, it's taking its
time to be built there eheh), but there are some unittest failures,
how do you want me to report them: one issue per arcihtecture? all of
them together? something else? I expect most of them will be wontfix,
but some might uncover hidden bugs

On Mon, Oct 5, 2015 at 5:32 AM, Jouni K. Seppänen <[hidden email]> wrote:
> Sandro Tosi <[hidden email]> writes:
>
>> in the meantime a new error came up, while building latex doc:
>>
>> ```
>> Adding blank page after the table of contents.
>> [1] [2] [1] [2]
>> Chapter 1.
>> (/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)
>>
>> ! Package inputenc Error: Unicode char \u8:́ not set up for use with LaTeX.
>>
>> See the inputenc package documentation for explanation.
>> Type  H <return>  for immediate help.
>>  ...
>>
>> l.163 Sté
>>            fan van der Walt's talk from SciPy2015
>
> The letter é in the source file is encoded using U+0301 COMBINING ACUTE
> ACCENT (UTF-8 bytes cc 81):
>
> $ git grep 'van der W' | hexdump -C
> 00000000  73 74 79 6c 65 5f 63 68  61 6e 67 65 73 2e 72 73  |style_changes.rs|
> 00000010  74 3a 53 74 65 cc 81 66  61 6e 20 76 61 6e 20 64  |t:Ste..fan van d|
> 00000020  65 72 20 57 61 6c 74 27  73 20 74 61 6c 6b 20 66  |er Walt's talk f|
> 00000030  72 6f 6d 20 53 63 69 50  79 32 30 31 35 0a        |rom SciPy2015.|
> 0000003e
>
> LaTeX would probably understand U+00E9 LATIN SMALL LETTER E WITH ACUTE
> better. I'll make a PR.
>
> --
> Jouni K. Seppänen
> http://www.iki.fi/jks
>
> _______________________________________________
> Matplotlib-devel mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-devel

--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

Sandro Tosi-4
sure here they are:
https://buildd.debian.org/status/package.php?p=matplotlib&suite=sid
(you will have to click on "all (1)" for each arch and then on the
Maybe-Succesful) if you skim thru some of them, lemme know which one I
can skip before submitting duplicate issues

On Tue, Oct 6, 2015 at 11:59 PM, Thomas Caswell <[hidden email]> wrote:

> Can you point me where the logs are?
>
> I suspect one issue per arch is the best approach.
>
> Tom
>
> On Tue, Oct 6, 2015 at 6:57 PM Sandro Tosi <[hidden email]> wrote:
>>
>> first of all, thanks a lot for the prompt support in making mpl ready
>> to be uploaded to Debian! which happened last night, and it is mostly
>> built on all our official architectures (except mips, it's taking its
>> time to be built there eheh), but there are some unittest failures,
>> how do you want me to report them: one issue per arcihtecture? all of
>> them together? something else? I expect most of them will be wontfix,
>> but some might uncover hidden bugs
>>
>> On Mon, Oct 5, 2015 at 5:32 AM, Jouni K. Seppänen <[hidden email]> wrote:
>> > Sandro Tosi <[hidden email]> writes:
>> >
>> >> in the meantime a new error came up, while building latex doc:
>> >>
>> >> ```
>> >> Adding blank page after the table of contents.
>> >> [1] [2] [1] [2]
>> >> Chapter 1.
>> >> (/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)
>> >>
>> >> ! Package inputenc Error: Unicode char \u8:́ not set up for use with
>> >> LaTeX.
>> >>
>> >> See the inputenc package documentation for explanation.
>> >> Type  H <return>  for immediate help.
>> >>  ...
>> >>
>> >> l.163 Sté
>> >>            fan van der Walt's talk from SciPy2015
>> >
>> > The letter é in the source file is encoded using U+0301 COMBINING ACUTE
>> > ACCENT (UTF-8 bytes cc 81):
>> >
>> > $ git grep 'van der W' | hexdump -C
>> > 00000000  73 74 79 6c 65 5f 63 68  61 6e 67 65 73 2e 72 73
>> > |style_changes.rs|
>> > 00000010  74 3a 53 74 65 cc 81 66  61 6e 20 76 61 6e 20 64  |t:Ste..fan
>> > van d|
>> > 00000020  65 72 20 57 61 6c 74 27  73 20 74 61 6c 6b 20 66  |er Walt's
>> > talk f|
>> > 00000030  72 6f 6d 20 53 63 69 50  79 32 30 31 35 0a        |rom
>> > SciPy2015.|
>> > 0000003e
>> >
>> > LaTeX would probably understand U+00E9 LATIN SMALL LETTER E WITH ACUTE
>> > better. I'll make a PR.
>> >
>> > --
>> > Jouni K. Seppänen
>> > http://www.iki.fi/jks
>> >
>> > _______________________________________________
>> > Matplotlib-devel mailing list
>> > [hidden email]
>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>>
>> --
>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>> My website: http://matrixhasu.altervista.org/
>> Me at Debian: http://wiki.debian.org/SandroTosi
>> _______________________________________________
>> Matplotlib-devel mailing list
>> [hidden email]
>> https://mail.python.org/mailman/listinfo/matplotlib-devel



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel
Reply | Threaded
Open this post in threaded view
|

Re: v1.5.0rc2

tcaswell
It looks like on arm and powerpc some of the spectral computations in mlap are returning complex instead of real values which is odd.

My guess on the sphinx issue is that the version of sphinx that is being spawned via a `Popen` call does not have the right PYTHONPATH and is trying to import a non-built version of matplotlib.

The tight-layout fail is one that we see intermittently on travis.

Those logs are burly. 

On Tue, Oct 6, 2015 at 7:02 PM Sandro Tosi <[hidden email]> wrote:
sure here they are:
https://buildd.debian.org/status/package.php?p=matplotlib&suite=sid
(you will have to click on "all (1)" for each arch and then on the
Maybe-Succesful) if you skim thru some of them, lemme know which one I
can skip before submitting duplicate issues

On Tue, Oct 6, 2015 at 11:59 PM, Thomas Caswell <[hidden email]> wrote:
> Can you point me where the logs are?
>
> I suspect one issue per arch is the best approach.
>
> Tom
>
> On Tue, Oct 6, 2015 at 6:57 PM Sandro Tosi <[hidden email]> wrote:
>>
>> first of all, thanks a lot for the prompt support in making mpl ready
>> to be uploaded to Debian! which happened last night, and it is mostly
>> built on all our official architectures (except mips, it's taking its
>> time to be built there eheh), but there are some unittest failures,
>> how do you want me to report them: one issue per arcihtecture? all of
>> them together? something else? I expect most of them will be wontfix,
>> but some might uncover hidden bugs
>>
>> On Mon, Oct 5, 2015 at 5:32 AM, Jouni K. Seppänen <[hidden email]> wrote:
>> > Sandro Tosi <[hidden email]> writes:
>> >
>> >> in the meantime a new error came up, while building latex doc:
>> >>
>> >> ```
>> >> Adding blank page after the table of contents.
>> >> [1] [2] [1] [2]
>> >> Chapter 1.
>> >> (/usr/share/texlive/texmf-dist/tex/latex/txfonts/t1txtt.fd)
>> >>
>> >> ! Package inputenc Error: Unicode char \u8:́ not set up for use with
>> >> LaTeX.
>> >>
>> >> See the inputenc package documentation for explanation.
>> >> Type  H <return>  for immediate help.
>> >>  ...
>> >>
>> >> l.163 Sté
>> >>            fan van der Walt's talk from SciPy2015
>> >
>> > The letter é in the source file is encoded using U+0301 COMBINING ACUTE
>> > ACCENT (UTF-8 bytes cc 81):
>> >
>> > $ git grep 'van der W' | hexdump -C
>> > 00000000  73 74 79 6c 65 5f 63 68  61 6e 67 65 73 2e 72 73
>> > |style_changes.rs|
>> > 00000010  74 3a 53 74 65 cc 81 66  61 6e 20 76 61 6e 20 64  |t:Ste..fan
>> > van d|
>> > 00000020  65 72 20 57 61 6c 74 27  73 20 74 61 6c 6b 20 66  |er Walt's
>> > talk f|
>> > 00000030  72 6f 6d 20 53 63 69 50  79 32 30 31 35 0a        |rom
>> > SciPy2015.|
>> > 0000003e
>> >
>> > LaTeX would probably understand U+00E9 LATIN SMALL LETTER E WITH ACUTE
>> > better. I'll make a PR.
>> >
>> > --
>> > Jouni K. Seppänen
>> > http://www.iki.fi/jks
>> >
>> > _______________________________________________
>> > Matplotlib-devel mailing list
>> > [hidden email]
>> > https://mail.python.org/mailman/listinfo/matplotlib-devel
>>
>> --
>> Sandro Tosi (aka morph, morpheus, matrixhasu)
>> My website: http://matrixhasu.altervista.org/
>> Me at Debian: http://wiki.debian.org/SandroTosi
>> _______________________________________________
>> Matplotlib-devel mailing list
>> [hidden email]
>> https://mail.python.org/mailman/listinfo/matplotlib-devel



--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi

_______________________________________________
Matplotlib-devel mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-devel