>> I'm working with someone  who is offering a qemu arm buildslave. The
>> build on the qemu arm hosts takes ~3 hours. I'm hesitant to have all of
>> the gerrit submissions take three hours to build. A daily build of the
>> git branches might be better in this case.
> 
> I'm already finding the cycle time of the IRIX build slave hugely
> frustrating, and would be opposed to adding anything that lengthens that
> time further.

I agree that this is frustrating.  However, the IRIX build time could be 
greatly reduced if parallel make worked properly; it can and does in other open 
source projects, but I believe that we're running into issues with Makefile 
assumptions about build order that don't necessarily hold true for IRIX.  It is 
a goal of mine to get this working, but I may need some guidance along the way. 
 Pointers welcome!

I might suggest that there are some types of source changes which may not 
require the full tree to be rebuilt ... I'm thinking of documentation changes 
in particular, but there may be other things in there as well?  In my view, the 
main advantage of the buildbot system is to test the actual code on as many 
platforms as possible.  If we were to optimize around that, for instance by 
having build submissions to documentation or WINNT subdirs, etc., go through 
different build processes, would that be better?  Do we need to test 
windows-only-related changes on UNIX?  Is buildbot configurable enough to do 
this?

Otherwise, a tiering system may be necessary, especially for very slow platforms

-Chaz

____________________________________________________________
Receive Notifications of Incoming Messages
Easily monitor multiple email accounts & access them with a click.
Visit http://www.inbox.com/notifier and check it out!
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel

Reply via email to