Inverted pulling of head/tags on latest builds

[davfsa]

Details

When fetching the branch/tags for latest builds, the method for fetching branches and tags is inversed, leading to failures when fetching them

Building a branch as latest
https://readthedocs.org/projects/hikari-py/builds/27320926/

Building a tag as latest
https://readthedocs.org/projects/hikari-py/builds/27320945/

Expected Result

Fetch succeeds

Actual Result

Fetch fails

[humitos]

Can you fill the template with the URL of you build, etc?

[davfsa]

Can you fill the template with the URL of you build, etc?

Meant to do that and then seem to have forgotten when adding the images, sorry!

Updated now @humitos!

[humitos]

This seems like a very basic case that it should work out of the box and that it should have been working for a long time. @stsewd did we change anything recently?

[davfsa]

This seems like a very basic case that it should work out of the box and that it should have been working for a long time

Funnily enough, this is something that has broken in the past (see
#10735). It seems to be (at least) a bit unexpected behaviour and not really something that is thought about a lot when implementing it (ref: #10735 (comment)).

But this seems like some boolean got inverted somewhere. It is weird that nobody else is complaining about it, since I would expect this to break in more cases

[davfsa]

Found out some more information on this issue.

I decided to check the builds for our page and it seems like latest is pulling correctly now. So it seems that updates coming from the branch receiving pushes are pulled properly, the issue comes when you update the default branch option (for reference, I am using the new portal at https://app.readthedocs.org/dashboard/hikari-py/edit/)

Screenshots of me trying this out. Failed builds are caused by me changing the value of the option and the successful ones are from pushes to the master branch. Also interesting to note, if I press the "Rebuild version" button, the build retries and succeeds this time (which is that build that is loading)

EDIT: This seems to only happen when switching from a tag to a branch. When switching from a branch to a branch or tag to tag, this doesn't happen

Hope this helps track the bug down, as I tried to have a look and I am stumped on what could be the code causing it

[stsewd]

Hi, this is indeed a problem when changing the default version from a branch to a tag or vice versa, when triggering a build we resync all versions and the state gets corrected.