Driving ARM secondary rebuild with script - builddep tree?

Nathanael D. Noblet nathanael at gnat.ca
Sun Apr 24 16:10:41 UTC 2011


On 04/24/2011 08:56 AM, Martin Langhoff wrote:
> While smarter people than me fix the boostrapping of F14 on ARM, I am
> looking at whether we can automate driving koji in an optimal order.
>
> I am looking at mass-rebuild.py from the releng scripts repo; and at
> yum-builddep.
>
>   - mass-rebuild gets a git checkout, bumps the rev, commits, pushes,
> we don't do that (!)
>
>   - mass-rebuild counts on tags and on koji metadata indicating that a
> pkg is blocked -- we don't do tags, and I don't know whether koji's
> "blocked" metadata indicates a builddep check
>
>   - if possible, we'd like to prioritize a particular set of packages
> that is in the critical path for testing on our hw -- for that, I was
> hoping to perform a recursive builddep tree, but it is... more complex
> than it seems at first blush. Is there any tool that performs
> recursive builddep checks, and defines an order (as yum does with deps
> when planning an install?)

Not sure if it'll do the ordering the way you want - however look at 
smock.pl... It works well for me to build a set of RPMS where one 
depends on the other. I made two modifications to the original smock.pl 
I found online, one allows for multiple arch builds independently, the 
other uses threads to build each arch independently. My personal testing 
using time had the build finish ~50% faster when using threads. (3-4min 
vs 7min). I provided upstream with the patches but I'm not sure if they 
were applied or not. Either way not sure if it will solve your problem. 
If it does and you need/want the threaded patches let me know.


-- 
Nathanael d. Noblet
t 403.875.4613


More information about the devel mailing list