[OE-core] [RFC][PATCH 0/2] WORKDIR and insane tweaks

Ross Burton ross.burton at intel.com
Tue Aug 13 10:10:58 UTC 2013


Hi,

Last night I was attempting a build with WORKDIR in a tmpfs to see what impact
it made on build time (the results were fairly interesting, I'll write them up
shortly).

I noticed that to move WORKDIR I had to replicate the full WORKDIR path logic,
which was a bit irritating but not that serious.  I then noticed that the
"rpaths" sanity test checks that WORKDIR is in fact TMPDIR/work, and does a
bb.fatal() if it isn't.  My best understanding here is that the check needs to
know the base of all possible WORKDIRS so that it can check for linkage directly
into another package's work directory.

Assuming that understanding is correct then both problems can be resolved with a
BASE_WORKDIR variable, which this series adds to bitbake.conf and then uses in
insane.bbclass.

Ross

The following changes since commit 16d522bcd1f1b7741577fa31fab7e2129da0cae9:

  maintainers.inc: reassign maintainers, remove obsolete recipes (2013-08-12 13:13:05 +0100)

are available in the git repository at:

  git://git.yoctoproject.org/poky-contrib ross/workdir

for you to fetch changes up to e91cbd17131101d1c470669acce66682ae06960a:

  insane: don't abort if workdir is not TMPDIR/work (2013-08-13 10:36:53 +0100)

----------------------------------------------------------------
Ross Burton (2):
      bitbake.conf: define WORKDIR in terms of BASE_WORKDIR
      insane: don't abort if workdir is not TMPDIR/work

 meta/classes/insane.bbclass |    5 +----
 meta/conf/bitbake.conf      |    3 ++-
 2 files changed, 3 insertions(+), 5 deletions(-)

Ross Burton (2):
  bitbake.conf: define WORKDIR in terms of BASE_WORKDIR
  insane: don't abort if workdir is not TMPDIR/work

 meta/classes/insane.bbclass |    5 +----
 meta/conf/bitbake.conf      |    3 ++-
 2 files changed, 3 insertions(+), 5 deletions(-)

-- 
1.7.10.4




More information about the Openembedded-core mailing list