Maintainer should be emailed when examples do not compile

[jnothman]

Unfortunately the HTML doc compilation does not return a non-zero exit status when examples don't compile. Perhaps this is a job for sphinx-gallery too (ping @Titan-C).

If docs fail to compile does maintainer currently get notified?

This behaviour would have avoided scikit-learn/scikit-learn#5170 and other example error issues.

[amueller]

no there is no notifcation currently.
Also, the docbuilder also doesn't upload (the current dev website is my manual build).

If we have notifications, it would be good to also run the non-plotting examples, which get checked even more rarely.

[amueller]

what does it mean for the docs to fail to compile? if sphinx exists with an error? There are usually many sphinx errors. If we included them in the travis build, we could probably avoid that. (ideas may be forming ^^).

The latexpdf is never build by an automatic service afaik.

[GaelVaroquaux]

One option would be to use CircleCI to build the docs, as done in nilearn: nilearn/nilearn#770 This raises a red flag in the github checks when examples do not build.

[Titan-C]

Comments on build process notification are welcome in
sphinx-gallery/sphinx-gallery#45

[jnothman]

Also, the docbuilder also doesn't upload (the current dev website is my manual build).

I thought we were using rackspace for something or other...? Do you want me to set up a cron job? It seems, indeed, that the current dev build is outdated (e.g. http://scikit-learn.org/dev/auto_examples/cross_decomposition/plot_compare_cross_decomposition.html is still broken).

[ogrisel]

We are already using a cron job, but it updates the sourceforge hosted version of the documentation, not the new github host. The script that is run by the cron job needs to be updated:

https://github.com/scikit-learn/sklearn-docbuilder/blob/master/srv/salt/update_doc.sh

[jnothman]

Got it. I assume as long as the ssh key used to upload to sourceforge also applies to an appropriate github account, that that is a fairly trivial change.

Also curious, does the cron output get output somewhere upon failure?

[amueller]

The change to the script is easy. The problem is more deploying the repo to
the rackspace host. I'm not really familiar with salt. But I guess ogrisel
is on it.
On Sep 2, 2015 02:59, "jnothman" [email protected] wrote:

Got it. I assume as long as the ssh key used to upload to sourceforge also
applies to an appropriate github account, that that is a fairly trivial
change.

Also curious, does the cron output get output somewhere upon failure?

—
Reply to this email directly or view it on GitHub
#3 (comment)
.