[PD-dev] buildbot on Macs and from git

[Hans-Christoph Steiner]

On Sep 21, 2010, at 6:57 PM, András Murányi wrote:

>
> I'm actually not going to have any time to look at this for a couple
> weeks, so please edit away.
>
> Ok, i will. I'm not very good at build building generally, and which  
> steps certain pd sources need, but i'll experiment. I'll bug you if  
> something fails and i really don't know why.
>
>
>
>
> I'll still be on email, so I can contribute here and there.  Yeah, I  
> was also thinking that the libs wouldn't stop the whole build.
>
> .hc
>
> Could you help me with these?
> - What is the "svn update" meant to do after rsync? Right now it  
> just fails...
>
> The rsync server is updated 2-3 times a day, the "svn update" is  
> meant to get any more recent changes.  I fixed that so it should  
> fail anymore.
>
>
> - Pdx "make package" fails on linux with "dpkg-deb: control  
> directory has bad permissions 700 (must be >=0755 and <=0775)", i  
> suppose it's easy to come around this, but i guess from the  
> Makefile...
>
> Yeah, this is annoying because builtbot sets the umask so that group  
> and other get no permissions.  but to build a .deb package you need  
> to have the files at least 0755 (user-rwx, group-rx, other-rx). I'm  
> not quite sure how to handle it.  My guess is that it would make  
> sense to set the umask on the buildbot builds to have group/other  
> read/execute perms.  I think that's umask 022.
>
>
> Seems that slaves can have their own umask set in /var/lib/buildbot/ 
> slavename/buildbot.tac
> Set 022 on my box, let's see if it helps.
> ps: cannot see because an older compile error came back: "cannot  
> find install-sh, install.sh, or shtool"

Arg, yes, stupid mistake, should be fixed.

>
> - Git update fails with "fatal: ambiguous argument '14159': unknown  
> revision or path not in the working tree. Use '--' to separate paths  
> from revisions". Is it a buldbot bug (tries to use svn version for  
> git?) or do you have another idea?
>
>
> Sounds like git and svn are fighting ;-) Sounds like the buildbot is  
> running git with the SVN revision number: "/usr/bin/git reset --hard  
> 14165".  I guess that happens when a new SVN checkin triggers a git  
> build.  I think we need to remove the SVN tracking for a  
> PBChangeSources style.
>
>
> Done. buildbot 0.8.1 has GitPoller() so one day we may be able to  
> set it up. Meanwhile we could take a look how this PBChangeSource()  
> can be triggered from the sourceforge git. One way is contrib/ 
> git_buildbot.py, another a commit script calling "buildbot  
> sendchange" and the last resort is parsing commit mails ("post- 
> receive-email" for git) with a MaildirSource() which is to be  
> written up. Knowing that GitPoller() already exists i'd recommend  
> the simplest solution but which one is that? :)
>
> Andras

The pollers are nice, but we can't use them because they assume that  
all builds are coming from that single git source.  If we want to  
setup multiple masters, then we could.  It sounds like the thing to do  
is investigate the PBChangeSource stuff to see if we can have the pure- 
data SVN and multiple git sources

.hc



----------------------------------------------------------------------------

"We have nothing to fear from love and commitment." - New York Senator  
Diane Savino, trying to convince the NY Senate to pass a gay marriage  
bill