On Wed, 2010-01-13 at 15:24 -0800, C Michael Sundius wrote:
> We've run into the problem where developers who are busily making
> changes to their code, building it, testing it and then doing that all
> over again and again are being bogged down because in one particular
> use case, several dependent objects (from different recipes/packages)
> are statically lined together into a single kernel module.
> Unfortunately when one recipe is rebuilt, the dependant recipe (that
> links all the modules together) is not:
> 
> if A depends on (the staging output of) B, and B is rebuilt, A is not
> (automatically rebuilt).
> 
> The conventional wisdom on the OE list is that "this is not supported
> though it would be nice to have"
> 
> In reading thru the bitbake-dev archive I came accross this posting:
> 
> https://lists.berlios.de/pipermail/bitbake-dev/2008-February/000385.html
> 
> in which Richard Purdie seems to have resolved this issue. though, I
> never saw a corresponding formal patch. 
> 
> 1) Did this code make it into Bitbake somehow?
> 2) If so how to I make use of it. 
> 3) if not does anyone have any comments on how to resolve this
> problem.

You might want to look at BB_STAMP_POLICY = "full" or maybe "whitelist".

These will mean that when any dependency changes a given package
rebuilds. "whitelist" mode uses BB_STAMP_WHITELIST to only handle full
dependencies on a subset of packages if I recall correctly.

> I'm trying to come up to speed on how bitbake works internally, but
> its slow going, if anyone could offer a hand to give me a clue where
> to start, I'd be eternally grateful. 

The above is handled in runqueue.py.

Cheers,

Richard

_______________________________________________
Bitbake-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/bitbake-dev

Reply via email to