Sage no longer working in Binder

[Eric Gourgoulhon]

Hi,

Apart from Sage's Docker image being still stuck to 9.3.beta8 (issue discussed at https://groups.google.com/g/sage-devel/c/fUpSEHwR5VQ/m/YVmsBSOyCAAJ), a new issue has appeared: one cannot run Sage at all in Binder, even with the 9.3.beta8 image. Every attempt fails at the "Launching server..." step, with the message

Launch attempt 1 failed, retrying...

This has been reported at ask.sagemath:

https://ask.sagemath.org/question/59187/binder-enivronment-using-sagemath-is-not-loading/

Eric.

[Dima Pasechnik]

isn't the issue on the Binder side?

Eric.

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/9e832ca7-4471-4a0c-968e-49594aafa9f3n%40googlegroups.com.

[Eric Gourgoulhon]

The full error message is

Launching server...

Launch attempt 1 failed, retrying...

Launch attempt 2 failed, retrying...
Launch attempt 3 failed, retrying...
Internal Server Error

[Eric Gourgoulhon]

The above error message has been obtained by clicking on the Binder button in the notebook https://nbviewer.jupyter.org/github/sagemanifolds/SageManifolds/blob/master/Notebooks/SM_basic_Schwarzschild.ipynb

[William Stein]

My understanding is that the Sage-9.3 and Sage-9.4 binaries are broken
on a large number of machines due to an issue with how openblas was
built. You probably have to use a sage-9.2 docker container, or wait
for sage 9.5 to come out.

<rant>Somebody probably should have deleted all sage-9.3 and 9.4
binaries from all download repos, announced widely about this problem,
and made an emergency 9.4.x release with the openblas fix in it and
nothing else. But since I'm not willing to do that work, who am I to
complain?</rant>

-- William

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/2c01e889-825b-469c-8c08-3412dd470f57n%40googlegroups.com.



--
William (http://wstein.org)

[Eric Gourgoulhon]

Le jeudi 30 septembre 2021 à 17:53:46 UTC+2, wst...@gmail.com a écrit :

My understanding is that the Sage-9.3 and Sage-9.4 binaries are broken
on a large number of machines due to an issue with how openblas was
built. You probably have to use a sage-9.2 docker container, or wait
for sage 9.5 to come out.

Thanks for your feedback. However, I am not sure this is the source of the issue here, because the Sage 9.3.beta8 Docker image used to work on Binder, until quite recently. So maybe there has been some change on the Binder side. I should try with Sage 9.2 though...

Eric.

[Eric Gourgoulhon]

Well, there is no need to test further against the Sage version since the user who reported the issue on ask.sagemath is using Sage 9.1:

https://github.com/calculuslab/Calculus_Lab/blob/main/Dockerfile

Eric. 

[Nathan Dunfield]

On Thursday, September 30, 2021 at 10:53:46 AM UTC-5 wrote:

My understanding is that the Sage-9.3 and Sage-9.4 binaries are broken
on a large number of machines due to an issue with how openblas was
built. You probably have to use a sage-9.2 docker container, or wait
for sage 9.5 to come out.

This is moot since Eric has reported OP is using 9.1, but at least on macOS both Jupyter and Sage start fine even with the openblas problem --- you need to import numpy (or similar) to get a crash.

Nathan

[enriqu...@gmail.com]

I have tried with sagemathinc/cocalc:latest and it works. There is a message of permission denied when running the notebooks, but it is possible to run them. Any other try with sagemath or sagemath-dev (installing jupyterlab) failed. This was my dockerfile:

FROM sagemathinc/cocalc:latest

COPY --chown=sage:sage . ${HOME}