[oe] Another race failure, not sure if this is a hard dependency or an inflight staging which races with another configure??

[Esben Haabendal]

On Thu, 2011-05-12 at 01:58 +0200, Leon Woestenberg wrote:
> Hello,
> 
> I think I found another build race here (must love multi-core machines).
> 
> perl-native is still being staged while file-native tries to run perl
> stuff resulting in "| perl: error while loading shared libraries:
> libperl.so: cannot open shared object file: No such file or
> directory".
> 
> Restarting bitbake continues the build (note that perl native sysroot
> staging finishes just after file-native failing).
> 
> Now, I have been thinking about these races; even though we might not
> have a dependency on perl, a configure stage may probe for perl, and
> if it exists, starts to run tests against it.
> So either perl must be present, or not at all. I.e. atomicly staged in sysroot.
> 
> Of course, could be this IS a hard dependency after all.
> 
> I'll investigate later, here is a small piece of log for your eyes.

I know I am kind of repeating myself, but these kind of problems do not
arise when using a per-recipe staging as I have proposed (and doing in
OE-lite).

I for one, always prefer to design out race-conditions instead of having
a continous (and always up-hill) battle with them.

/Esben