[OE-core] Package recipes change proposal (system-wide) - name wise

Joshua Lock josh at linux.intel.com
Thu Jul 26 20:44:43 UTC 2012


On Thu, 2012-07-26 at 20:40 +0100, Richard Purdie wrote:
> On Wed, 2012-07-25 at 07:23 +0000, Iorga, Cristian wrote:
> > For me, having the version number contained in the name of package recipe is a little bit puzzling.
> > For example: libpcre/libpcre_8.31.bb
> > 
> > Is there a solid technical reason to not have it like this?
> > 
> > libpcre/libpcre.bb
> > and inside the recipe to have a PV variable defined:
> > PV="8.31"
> 
> FWIW, you can do that and it will just work, even today. Its just not
> the convention we've "grown up" with.
> 
> > In that case, the upgrade/update process would not involve performing a "git move" operation.
> > In my opinion, this "git move" operation is something to be avoided, as it puzzles a little bit the versioning system and complicates the review process.
> > 
> > Of course, some changes in the bitbake system would be involved, but I guess would not be too complicated.
> > Also, there will be a volume of work to be performed for changing the name of recipes and adding the PV variable inside the recipe.
> > But I guess that a script could solve that, followed by some manual review.
> 
> The original idea was that updating to new versions was easy, it was a
> mv operation. The checksums have of course complicated this idea but the
> principle still applies. It also lets you see versions without having to
> view the files themselves which I know I personally find very useful.
> 
> Also, as others have mentioned, git can detect move operations if you
> tell it to.

FYI the create-pull-request script passes -M40 to git format-patch,
which tells git to:

"should consider a delete/add pair to be a rename if more than 40% of
the file hasn’t changed."

Though with the size of some recipes perhaps the -M value should be
increased, or not passed at all?

Joshua





More information about the Openembedded-core mailing list