[Buildbot-devel] [ANN] Buildbot-0.7.10

[Marcus Lindblom]

Almad wrote:
> On Friday 06 of March 2009 11:07:39 Marcus Lindblom wrote:
>> Hi,
>>
>> It was changed (in reponse to another ticket) since a full clone is
>> expensive to do over and over again. Rather, we just do 'hg init' if we
>> need to (inital setup + error recovery and maybe branch shifts), then
>> 'hg pull' from thereon, to make things leaner and speedier.
> 
> hg init + hg pull is different from hg clone?

Only the sense that hg clone adds the clone URL as default path to 
.hg/hgrc, which we discovered was critical for some users.

>> I suppose the first checkout could be a 'hg clone' instead, followed by
>> the same 'hg pull' as now.
> 
> That is for "update" mode, right? But what about "export"?

For export, no .hg (or .git) dir exists, so it's a non-issue there. I 
suppose 'hg archive' should be used here too (just specify a remote 
repository, instead of a local as expected when doing 'copy').

> Actually, I'm experiencing similar problems on our git project; problem 
> especially is, that git init + git fetch do not fetch tags, so I have to 
> manually add git fetch -t.

IIRC, most DVCS:es has the same problem, since the tag is created in a 
different revision than what the tag actuallly tags. I attempted to 
address this for Mercurial by always pulling the entire repo, thus 
ensuring I get the revision where the tag was comitted, i.e. the tag itself.

> I think that in respect to new VCS around, it will be nice to create few more 
> commands and address differences.

I'm not aware of use-cases here (being quite new to DVCS:es). What new 
commands are you thinking of?

> (Well, #276 and #263 on my plate, so I might try during PyCon ,))

Neat. :)

Creating tickets for each VCS where this applies sounds like a good idea 
(i.e. #276 commment). It helps ppl quickly gauging the state of support 
BB has for a certain VCS.

Cheers,
/Marcus