[Buildbot-commits] [Buildbot] #97: Use svn switch to improve efficiency

[Buildbot trac]

#97: Use svn switch to improve efficiency
------------------------+-----------------------
Reporter:  retracile    |       Owner:
    Type:  enhancement  |      Status:  reopened
Priority:  major        |   Milestone:  0.8.+
 Version:  0.8.3        |  Resolution:
Keywords:  svn, sprint  |
------------------------+-----------------------

Comment (by dcoshea):

 Replying to [comment:1 dustin]:
 > You write:
 >  An 'svn switch' of a working copy containing unversioned files can in
 some cases fail; that is not handled in this patch.
 > which is what worries me.  I'm always in favor of burning a few more
 cycles over introducing spurious build failures.

 I note that in 0.8.6p1, a slave-side 'svn update' can "silently" fail if
 there are unversioned files in the working copy which were added in the
 repository - I've seen the following appear in the log for the SVN step:

 {{{
 svn: warning: Failed to add file '[...]': an unversioned file of the same
 name already exists
 }}}

 but the step was marked as passed, resulting in hard-to-diagnose build
 failures later.

 It seems to me, then, that doing an 'svn switch' isn't more risky than
 doing an 'svn update', and in either case, telling the step to do a clean
 checkout every time is a way to avoid these problems.


 Replying to [comment:5 dustin]:
 [...]
 > I'd rather that be "switch_branches" or something like that.  You're
 right that it applies to both modes, and multiple behaviors.
 >
 > >  1. I found a `Source` in `steps/source/oldsource.py`... is that the
 one that the master-side `SVN` inherits from?
 >
 > Yes.

 Doesn't "oldsource" mean "slave side"?

-- 
Ticket URL: <http://trac.buildbot.net/ticket/97#comment:10>
Buildbot <http://buildbot.net/>
Buildbot: build/test automation