Hi Ross,
Thanks for your reply, please see my comments inline.
On 02/27/2014 09:05 PM, Burton, Ross wrote:
Hi Robert,
Having a look at this and playing with my own toy archiver makes me
think that this shouldn't be using the postfuncs at all.
Yes, I gree as we had talked.
The main task, deploy_archives, should depend on the subtasks that it
needs and be simply ordered before do_build. I see that your code
does this but changes it to be before do_rootfs if the recipe is an
image: I'm not sure I understand why this is as do_rootfs is just a
task that executes before do_build so this shouldn't need
special-casing.
This part of code is derived from the previous archiver.bbclass, I will
remove it.
Why did you implement the subtasks as postfuncs instead of normal
tasks? This has the effect of causing more work to be done if the
What I had thought was that use the postfuncs will always make sure it
will run, for example, if do_unpack runs, then do_unpack[postfuncs] will
always run, but if we use a new task:
addtask do_ar_unpacked after do_unpack before do_patch
If do_patch doesn't run, the do_ar_unpacked would not run either, as had
talked, we will use new ${S}, so this problem will be gone, I will use
new tasks.
archiver needs to run: for example do_ar_unpacked() is a postfunc on
do_unpack so either enabling/disabling the archiver or editing the
archiver class causes the do_unpack hash to change, so everything
afterwards (do_patch, do_configure, do_compile, ...) has to
re-execute.
My toy archiver was implemented as a new task which is after do_fetch
but before do_build. This means that it executes after fetch has
completed but is outside of the
fetch->patch->configure->compile->install sequence.
https://github.com/rossburton/meta-ross/blob/master/classes/rearchiver.bbclass
(it also has a way of archiving the original sources in a way that
actually archives the original sources)
Great, thanks, I will send a V3 sooner.
// Robert
Ross
_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.openembedded.org/mailman/listinfo/openembedded-core
