Separate update and checkout steps

[stsewd]

So, I can see that the update commands of the other VCS' do the same that the first part of checkout. And we don't use update in any part of the code.

Closes #4259

And half fixes (#4890), right now users without a master branch are blocked and can't do anything to build their docs (previously users could put any text on the default branch, but now it's a select field).

With this change, the first build will fail, but now they can select another version as default branch.

Still wip, a lot tests will break probably and I neeed to do the same with the other VCS'

TODO

• git
• svn
• hg
• bzr

[codecov]

Codecov Report

Merging #4901 into master will increase coverage by 0.1%.
The diff coverage is 76.31%.

@@            Coverage Diff            @@
##           master    #4901     +/-   ##
=========================================
+ Coverage   76.65%   76.76%   +0.1%     
=========================================
  Files         158      158             
  Lines       10057    10052      -5     
  Branches     1269     1267      -2     
=========================================
+ Hits         7709     7716      +7     
+ Misses       2007     1995     -12     
  Partials      341      341

Impacted Files	Coverage Δ
readthedocs/vcs_support/backends/git.py	80.83% <100%> (+0.48%)	⬆️
readthedocs/vcs_support/backends/hg.py	64.51% <100%> (+5.42%)	⬆️
readthedocs/vcs_support/backends/svn.py	29.5% <25%> (+1.81%)	⬆️
readthedocs/vcs_support/backends/bzr.py	32.72% <50%> (+0.58%)	⬆️
readthedocs/projects/tasks.py	71.2% <77.77%> (+0.17%)	⬆️
readthedocs/restapi/client.py	87.87% <0%> (+0.78%)	⬆️
readthedocs/vcs_support/base.py	90.56% <0%> (+1.88%)	⬆️

[stsewd]

Diff coverage is falling because we don't have tests for bzr or svn. Otherwise, this is good to merge.

[agjohnson]

The code here looks good! I'm still not 100% on the potential fallout from this though, so it might be good to wait on a 2nd review

[agjohnson]

@humitos you're probably most familiar with this code, marking you for another review of this

[ericholscher]

Agreed this makes sense.

[humitos]

This makes sense to me.

Now, it's cleaner than a checkout only means checkout instead of "update and checkout" or similar.

I don't see any potential problem with this on this flow, but we should check if there are other places where the VCS classes are used and confirm that we don't need to adapt those flows.

[humitos]

I think the comment about hitting this endpoint could trigger a new build for the stable is still relevant here, right?

[stsewd]

It was moved to the function's docstring https://github.com/rtfd/readthedocs.org/pull/4901/files?utf8=%E2%9C%93&diff=unified#diff-b9399e1d3499066c5564f98a620e8881R162

[humitos]

You lost some info when moving the docstring. It talks about the stable version.

In fact, it may worth to expand it a little more.

[stsewd]

Done

[stsewd]

but we should check if there are other places where the VCS classes are used and confirm that we don't need to adapt those flows.

I grep the project and didn't find any additional checkout calls

[stsewd]

I'm merging this to start testing it more locally before the release.