[Debian-science-sagemath] Getting sagemath back into testing

[Tobias Hansen]

On 11/9/20 11:32 AM, Tobias Hansen wrote:
> On 11/9/20 11:28 AM, Dima Pasechnik wrote:
>> On Mon, Nov 9, 2020 at 11:18 AM Dima Pasechnik <dimpase at gmail.com> wrote:
>>> On Mon, Nov 9, 2020 at 11:06 AM Julien Puydt <julien.puydt at gmail.com> wrote:
>>>> Le dimanche 01 novembre 2020 à 10:22 +0000, Tobias Hansen a écrit :
>>>>> And three of the patches in sagemath and many of the failing doctests
>>>>> are due to our ancient ipywidgets version...
>>>> I don't remember if I already answered to that point, but I
>>>> investigated the matter :
>>>>
>>>> - ipywidgets has a python part and a javascript part ; the python part
>>>> looks ok, but for the javascript part, it looks like it wants
>>>> jupyterlab ;
>>> Sage 9.2 does not need jupyterlab, so I'm not sure why this would come up.
>>> https://github.com/jupyter/notebook/releases/tag/6.1.4
>>> does not depend on jupyterlab AFAIK
>>> (Sage is on 6.1.1, but it's a minor difference)
>>>
>> as far as ipywidgets are concerned, Sage ships the latest version,
>> 7.5.1, and again, it does not need jupyterlab
>> Here are their deps:
>>
>> $ cat build/pkgs/notebook/dependencies
>> $(PYTHON) | $(PYTHON_TOOLCHAIN) ipython jupyter_client ipykernel
>> nbconvert nbformat jinja2 tornado terminado send2trash
>> prometheus_client
>>
>> $ cat build/pkgs/ipywidgets/dependencies
>> $(PYTHON) widgetsnbextension | $(PYTHON_TOOLCHAIN)
>>
> I think the difference is that sagemath does not really build the javascript files of ipywidgets from their true source, but just installs the distributed files. In Debian we do build them with some really ugly scripts to satisfy Debian policy and because of this it became really hard to update this package.
>
> Best,
>
> Tobias
>
Another part of the problem is unbundling javascript dependencies. I'm not sure about the details though..

Best,

Tobias