[Buildbot-devel] Behavior of mode=copy

[bugs at almad.net]

On Mon, 12 May 2008 23:06:10 -0700, Stephen Davis <buildbot at soundgeek.org> wrote:

> On May 12, 2008, at 7:25 AM, bugs at almad.net wrote:

>

> While apparently more important for VCs that keep the entire

> repository history locally, a VC-aware "copy" operation would benefit

> CVS, SVN, etc. as well by avoiding each VC's specific administrative

> directories.



Yes, I was just unsure about each and every VCs implementing this.





>> Mercurial has build-in command archive, which just copies active

>> revision to another directory, which I think is more usable for this

>> case (and for my case speeds up update from 1 hour to 3 minutes).

>> I've attached my patch to #263.

>

> I'd think that having the VC-specific code implement the copy

> operation would be the right way to do this.  For example, a

> subversion mode would be like "cp -rp" but it would exclude the .svn

> folders.



As above.



> That's probably more than you were planning on tackling but perhaps

> you could file an enhancement request so that all VCs benefit.



OK, I'll fill it.



Main question was if I should modify test cases to change expected behaviour (= if it's not required to copy modified behaviour).



If we're decide to be vague, it's (or should be :)) then easy to modify test to add+commit unversioned files, use cp -rp as default and override doCopy with VC-specific behaviour.



My main concer was that that in spirit of testing corner cases, we should add a test that is ensuring that unversioned files are *not* copied. This will grant unambiguous behavior, but will broke all VCs tests until fixed + I don't know how this could be implemented if VC would not implement it.



Anyway, I have filled #276, test-suite decision and implementation for other VCs is up to you, developers ,)



> stephen



Almad