[poky] Broken dependency behavior in master?

[Richard Purdie]

On Mon, 2011-04-04 at 17:32 +0100, Paul Eggleton wrote:
> On Friday 01 April 2011 19:49:13 Scott Garman wrote:
> > I've reproduced this now, multiple times, using builds from scratch,
> > including the fact that reverting that commit fixes the the problem for me.
> 
> I can confirm this, even though on the face of it the patch should not cause 
> any issues. 
> 
> I haven't completely isolated the cause yet, however the problem occurs in  
> sstate_task_postfunc -> sstate_install -> docbook_sgml_dtd_sstate_postinst 
> which oddly seems to be being merged into the do_populate_lic task. I'm not 
> sure but I'm assuming that these sstate functions should not be being executed 
> under these circumstances.

Some data points on this:

"bitbake docbook-sgml-dtd-3.1-native -c populate_lic"

is the reproducer and the problem is the line:

SSTATEPOSTINSTFUNCS += "docbook_sgml_dtd_sstate_postinst"

in docbook-sgml-dtd-native.inc. This is running at populate_lic time as
well as populate_sysroot when its meant to run.

The commit Scott found is just a coincidence on some task order and not
responsible for this issue. I believe Scott is going to test+send a fix
which ensures the postinst only runs for the populate_sysroot postinst.

Cheers,

Richard