CI: Documentation built but not published in PRs

[datapythonista]

I was taking a look at the build of the documentation, and it looks like in every PR it is built, but it's not being published:

From https://github.com/pandas-dev/pandas-docs-travis
 * [new branch]      gh-pages   -> origin/gh-pages
origin	https://@github.com/pandas-dev/pandas-docs-travis.git (fetch)
origin	https://@github.com/pandas-dev/pandas-docs-travis.git (push)
remote: Anonymous access to pandas-dev/pandas-docs-travis.git denied.
fatal: Authentication failed for 'https://@github.com/pandas-dev/pandas-docs-travis.git/'

On commits to master, looks like it's working as expected (we should probably update the url):

From https://github.com/pandas-dev/pandas-docs-travis
 * [new branch]      gh-pages   -> origin/gh-pages
origin	https://[secure]@github.com/pandas-dev/pandas-docs-travis.git (fetch)
origin	https://[secure]@github.com/pandas-dev/pandas-docs-travis.git (push)
remote: This repository moved. Please use the new location:        
remote:   https://github.com/pandas-docs/pandas-docs-travis.git        
To https://github.com/pandas-dev/pandas-docs-travis.git
 + 6f622be...de98643 gh-pages -> gh-pages (forced update)

Building the docs takes around 45 minutes, so I think we should probably build only when merging the PRs, and save some travis resources? Not sure how to do it, but I guess there is a way inside travis to know if we're in a PR or in a commit to master.

May be in the future we can consider publishing the docs for each PR, but that would require some development, as fixing the error here would publish all PRs into the same place, which is not very useful.

The relevant files are .travis.yml and ci/build_docs.sh.

CC: @TomAugspurger @jorisvandenbossche

[TomAugspurger]

I like building on every PR, since (in theory) we'll be able to fail documentation builds when something is incorrect #22743 Scikit-Learn has a way to upload the built docs on every PR to a unique URL: #17921

…

On Tue, Nov 6, 2018 at 4:01 PM Marc Garcia ***@***.***> wrote: I was taking a look at the build of the documentation, and it looks like in every PR it is built, but it's not being published: From https://github.com/pandas-dev/pandas-docs-travis * [new branch] gh-pages -> origin/gh-pages origin ***@***.***/pandas-dev/pandas-docs-travis.git (fetch) origin ***@***.***/pandas-dev/pandas-docs-travis.git (push) remote: Anonymous access to pandas-dev/pandas-docs-travis.git denied. fatal: Authentication failed for ***@***.***/pandas-dev/pandas-docs-travis.git/' On commits to master, looks like it's working as expected (we should probably update the url): From https://github.com/pandas-dev/pandas-docs-travis * [new branch] gh-pages -> origin/gh-pages origin ***@***.***/pandas-dev/pandas-docs-travis.git (fetch) origin ***@***.***/pandas-dev/pandas-docs-travis.git (push) remote: This repository moved. Please use the new location: remote: https://github.com/pandas-docs/pandas-docs-travis.git To https://github.com/pandas-dev/pandas-docs-travis.git + 6f622be...de98643 gh-pages -> gh-pages (forced update) Building the docs takes around 45 minutes, so I think we should probably build only when merging the PRs, and save some travis resources? Not sure how to do it, but I guess there is a way inside travis to know if we're in a PR or in a commit to master. May be in the future we can consider publishing the docs for each PR, but that would require some development, as fixing the error here would publish all PRs into the same place, which is not very useful. CC: @TomAugspurger <https://github.com/TomAugspurger> @jorisvandenbossche <https://github.com/jorisvandenbossche> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#23536>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIufFMbE_5PfOC-84l5ihHmPBXKODks5usgajgaJpZM4YRW4Q> .

[datapythonista]

Oh, sorry, I missed that PR. That makes sense, I didn't think about it.

Then, the first thing would be to move the docs outside of the allowed failures in travis, which I don't see discussed or implemented there, right?

[jorisvandenbossche]

Yes, we should IMO for sure build the docs also in PRs. I currently often look in the log to see if something in the PR is going wrong regarding the docs, and automatically failing there as @TomAugspurger is trying to do would be a big help (now we are constantly introducing mistakes that needs to be corrected afterwards)

I think we already mentioned that a few times in various places, but without taking action, that we should indeed move it outside of the allowed failures.

[datapythonista]

Makes sense. Closing, as #22743 and #17921 already cover all what needs to be done.