[OE-core] Annoyances with dylan

Richard Purdie richard.purdie at linuxfoundation.org
Sat Apr 27 10:19:15 UTC 2013


On Fri, 2013-04-26 at 16:47 +0200, Koen Kooi wrote:
> Hi,
> 
> I have 2 big annoyances with dylan:
> 
> 1) Multimachine builds trigger spurious warnings:
> 
> WARNING: The recipe bonescript is trying to install files into a shared area when those files already exist. Those files and their manifest location are:
>    /build/v2013.06/deploy/eglibc/licenses/bonescript/LICENSE
>    /build/v2013.06/deploy/eglibc/licenses/bonescript/generic_MIT
> 
> I get that every time after switching machines for each recipe/license.

This sounds like a symptom of a deeper problem. At a guess bonescript
has some machine specific element in it but its not marked as machine
specific and hence the system is getting confused. Do you have a list of
the recipes/licenses that are showing this? It is just license issues?

> 2) PRSERV makes things go backwards:
> 
> ERROR: Package version for package gadget-init-network went backwards which would break package feeds from (0:1.0-r19.3 to 0:1.0-r19.0)
> 
> To trigger 2) I updated meta-linaro and switches meta-minnow to the
> new rebased branch, delete $TMPDIR (prserv db is outside out TMPDIR)
> and started building images. I only happens for some recipes, not all.
> I can't deduce a pattern or easily reproduce it. Bug filed some months
> ago.

IIRC we looked into that bug quite extensively and couldn't find a way
to reproduce it. If we want to get it fixed, we're going to need some
help distilling it down to a simpler test case to reproduce so we can
understand the issue, then we can look at fixing it.

> On the flip side: *ZERO* problems with sstate so far :)

Nice to know :)

Cheers,

Richard





More information about the Openembedded-core mailing list