Re: Reproducible builds (Was Re: Checksums in Bitbake)

[Sander van Grieken <sander@3v8.net>]

On Saturday 27 March 2010 19:31:49 GNUtoo wrote:
> On Thu, 2010-03-25 at 08:45 +0100, Frans Meulenbroeks wrote:
> > I've seen too often (also outside OE) that two engineers take the same
> > source yet still get different results, and that bugs at a customer
> > site cannot be reproduced in the lab (and yes, I do know there are
> > other ways to tackle this problem) 
> Also:
> bitbake optionaldep
> bitbake package
> And:
> bitbake package
> 
> could result in different binaries/packages due to configure picking
> optionaldep in the first case and not in the second one.

So, building any of the *-image recipes is non-deterministic, if it pulls in optional deps 
for other recipes? It then comes down to the build-queue ordering, which is probably not 
guaranteed to have a predictable order, especially with a high number of parallel bitbake 
threads.

Sounds like there's a need for something similar to Gentoo's USE flags.

grtz,
Sander