named branches in mercurial step are not supported

[dsallings]

Named branches are currently not supported in the Mercurial source step.

The attached patch models hgbuildbot's inrepo vs. dirname branches and allows the user to select the branching method.

[dsallings]

allow named branches in mercurial step

[dustin]

Looks good. Want to add a test case and some documentation?

[warner]

yay! I've love to commit this. But yeah, we need tests and docs. The test case should be in buildbot/test/test_vc.py, with a MercurialBranchInRepoHelper and MercurialBranchInRepo classes (as subclasses of their non-branchinrepo cousins). The MercurialBranchInRepo class needs to add these branchtype= arguments, and the helper needs to construct a repository that contains the extra branches in-repo instead of as separate repos.

thanks!

[marcusl]

I'm working on a test but I'm only halfway there yet. I've run into some problems with the auto-clobbering detection and I'm not really sure how to best tackle it.

I'll post a patch with my current progress, and see if anyone else could follow up.

[marcusl]

MercurialInrepoTest? + some random fixes on mercurial source step

[marcusl]

I tried to add some manual code for detecting stale files and return an error (to trigger the clobber step), but that's likely not the right way to do it.

One should probably leave that out completely and test for branches instead?

[marcusl]

MercurialInRepoTest? (new, test only)

[marcusl]

mercurial source step changes (new)

[marcusl]

These two last patches seem to make things run properly. Please ignore the 'mercurial_inrepo_test_187.diff' posted previosly.

(How does one remove attached files?)

It's 2.00 am so things might not be very pretty (especially as I fall back to a complete re-run of vcfull and vcupdate, not knowing how to do that except with some small wrapper functions), but at least all the tests (except tryPatch, see #400) complete successfully, for both Mercurial and MercurialInRepo?.

I've only tested this with hg 1.0.2, on Vista 64bit.

[marcusl]

I'm planning on moving all patches to  http://github.com/marcusl/buildbot-187/tree/master to make the merge easier, but its w.i.p. at the moment.

[marcusl]

Ok. I've put the patches there. Marking for review.

[marcusl]

Correction: There reference to #400 should be #398. (#400 still prevents the tryPatch tests to run successfully).

Sorry about the confiusion.

[marcusl]

Re-did the clone and stored the fixes on a separate branch instead (per SubmittingPatches).

See  http://github.com/marcusl/buildbot/tree/bug187 instead.

[dustin]

hmm, buildbot.test.test_vc.Mercurial.testCheckoutHTTP hangs for me. Any idea why?

[marcusl]

Hmm. I did run it successfully before, but it has occasionally borked. (claiming a 'dirty' or 'unclean' reactor, which doesn't say much. Often, just re-running only the HTTP-test helped.)

Anyway, I'll double-check if there was something that didn't get through to the git-repo (I worked in a darcs-clone previously and I tried to clean things up a bit when doing the git-patch).

[marcusl]

It doens't work here either, but if I enable deferred-debugging in test_vc.py the test passes.

#defer.Deferred.debug = True

That's what probably made it work before. Ideas?

[marcusl]

Actually, it seems to work every other time, give or take, regardless of what I do.

[marcusl]

I've added an extra deffered causing testCheckoutHTTP to wait until 'hg serve' has completed. This fixes the problem on my end.

This has been pushed to the bug187 branch on my git clone.

[dustin]

mreged -- thanks for hanging in there!