Re: SetScene tasks hang forever?

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Wed, 2012-05-09 at 16:04 -0700, Rich Pixley wrote:
> On 5/9/12 12:52 , Richard Purdie wrote:
> > On Wed, 2012-05-09 at 10:51 -0700, Rich Pixley wrote:
> >> On 5/8/12 05:34 , Richard Purdie wrote:
> >>> On Sun, 2012-05-06 at 10:36 -0700, Rich Pixley wrote:
> >>>> Any other ideas?
> >>> Well, this clearly doesn't happen with master or in any combination of
> >>> the layers most users are using. The logical conclusion would be that
> >>> there is something in your layer that is somehow triggering this.
> >> No private layer involved.
> >>
> >> I do have a makefile which encapsulates the environment stuff, but
> >> that's it.
> >>> Of course since that layer is secret and you can't show us it, we have a
> >>> bit of a problem. Can you reproduce the bug against public code?
> >> Done.  (Our layer is becoming open, we're committed to it, but it's a
> >> long process internally).
> >>> Are you by any chance setting BB_STAMP_POLICY somewhere?
> >> Yes.  BB_STAMP_POLICY = "full".
> >>
> >> I'll attach a copy of my local.conf and bblayers.conf.
> > I'm 95% sure its BB_STAMP_POLICY = "full" causing the problems. The idea
> > is really that sstate and other recent developments obsolete the "full"
> > stamp code. I'm not sure it actually gets on with the setscene stamps
> > the sstate code generates, as I suspect you're discovering.
> >
> > We could try and fix the "full" policy, or we could just remove it.
> > Looking at the code for the function that deals with this in
> > runqueue.py, I can see where problems could occur.
> >
> > So I guess I'm asking if we should fix that or can we remove it?
> Um... I'm not sure.
> 
> In the past, that was required to get everything built in an incremental 
> fashion.  That is, if A depended on B depended on C and C changed, 
> BB_STAMP_POLICY = "full" was the only way to get A to be rebuilt 
> automatically.
> 
> Are you saying that this happens automatically now even without the 
> BB_STAMP_POLICY = "full" setting?  Or that some other setting is more 
> appropriate and perhaps has semantics I don't know?  Or that the current 
> default is no incremental rebuilds?  Or... ?

The settings that are now recommended are:

BB_SIGNATURE_HANDLER ?= 'OEBasicHash'
OELAYOUT_ABI = "8"

This requires a rebuild since the stamp file format changes, hence the
ABI number increase. Currently, poky and angstrom use these settings
amongst others but its not default in OE-Core. I'll likely propose a
change to make it the default soon though.

This would then make BB_STAMP_POLICY = "full" obsolete and yet
incremental builds will work correctly and likely rebuild less things
(only really what potentially changed).

Cheers,

Richard