On Wed, 11 Mar 2015, Richard Purdie wrote:
On Wed, 2015-03-11 at 12:22 -0500, Mark Hatle wrote:
On 3/11/15 11:30 AM, Robert P. J. Day wrote:
i recall that richard purdie might have explained this once but i
have no idea where that post is and i need to add something to it,
anyway ... how many ways are there to delete a task from a recipe,
and what are the differences?
so far, i've seen:
1) bitbake supports a deltask directive
The task and all it's dependencies go away. So anything dependent on it
and the
things it depends on, are no longer dependent.
2) you can use the [noexec] task flag
The task and it's dependencies remain, but nothing is executed.
3) i've also seen redefining the task with a no-op : command
The task is executed and does nothing. (I believe this dumps a stamp file
and
the noexec does not.. but I may be wrong.)
What Mark said. 3) is the old way of doing 2), we added 2) to make
it more efficient since there is no task execution overhead (which
at one point was measurable, it may be less so now).
so that first part -- that 2) is a newer, tidier(?) way of doing 3)
-- is pretty much what i remember from once upon a time. good.
as for 1), i'm still confused by that explanation. say we have the
dependency chain A - B - C ... A depends on B which depends on C. if
i deltask B, then any dependency links which involve B will
disappear, both incoming and outgoing in the DAG. but will A be
adjusted to now depend on C? (i'm guessing not, just want to be sure.)
finally, is it just a matter of personal taste whether one deletes a
task using deltask or uses [noexec]? i'm looking at examples under
OE and here are a couple.
clearly, for image.bbclass, there are a *pile* of tasks that are not
relevant, so the following makes sense:
meta/classes/image.bbclass:do_fetch[noexec] = 1
meta/classes/image.bbclass:do_unpack[noexec] = 1
meta/classes/image.bbclass:do_patch[noexec] = 1
meta/classes/image.bbclass:do_configure[noexec] = 1
meta/classes/image.bbclass:do_compile[noexec] = 1
meta/classes/image.bbclass:do_install[noexec] = 1
meta/classes/image.bbclass:do_populate_sysroot[noexec] = 1
meta/classes/image.bbclass:do_package[noexec] = 1
meta/classes/image.bbclass:do_package_qa[noexec] = 1
meta/classes/image.bbclass:do_packagedata[noexec] = 1
meta/classes/image.bbclass:do_package_write_ipk[noexec] = 1
meta/classes/image.bbclass:do_package_write_deb[noexec] = 1
meta/classes/image.bbclass:do_package_write_rpm[noexec] = 1
meta/classes/image.bbclass:do_bundle_initramfs[noexec] = 1
similarly, for native.bbclass, the following also makes sense, but
uses deltask:
meta/classes/native.bbclass:deltask package
meta/classes/native.bbclass:deltask packagedata
meta/classes/native.bbclass:deltask package_qa
meta/classes/native.bbclass:deltask package_write_ipk
meta/classes/native.bbclass:deltask package_write_deb
meta/classes/native.bbclass:deltask package_write_rpm
meta/classes/native.bbclass:deltask package_write
so why does the first use [noexec] and the second use deltask? is
there a functional difference?
rday
p.s. i believe i appreciate the difference in *mechanics* of the
above. [noexec] doesn't affect the DAG of task dependencies, it simply
turns off the execution of that recipe-specific task, while deltask
affects the actual underlying task DAG.
--
Robert P. J. Day Ottawa, Ontario, CANADA
http://crashcourse.ca
Twitter: http://twitter.com/rpjday
LinkedIn: http://ca.linkedin.com/in/rpjday
--
___
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core