]> code.ossystems Code Review - openembedded-core.git/commit
uninative: Don't use single sstate for pseudo-native
authorRichard Purdie <richard.purdie@linuxfoundation.org>
Sat, 21 Nov 2020 16:16:40 +0000 (16:16 +0000)
committerSteve Sakoman <steve@sakoman.com>
Thu, 28 Jan 2021 14:41:47 +0000 (04:41 -1000)
commit7c5881cf9baa7ea9fcfed382f24334cf8f96f9c2
tree41ab54fa47b492ed6927eee41b8830e486c27984
parent3a826c352c9c4f24fd2b3a4e67f60ff3496b4e0f
uninative: Don't use single sstate for pseudo-native

pseudo-native is a bit special. It conditionally compiles in support for
xattr, statx and statvfs amongst other options. If a pseudo-native binary is
used on a system where these functions are present but it wasn't compiled in
we see hard to debug permissions problems.

An example is the devtool.DevtoolExtractTests.test_devtool_deploy_target
oe-selftest which shows a cryptic error:

  File "/home/pokybuild/yocto-worker/oe-selftest-fedora/build/meta/lib/oeqa/selftest/cases/devtool.py", line 1388, in test_devtool_deploy_target
    self.assertEqual(filelist1, filelist2)
  File "/usr/lib64/python3.9/unittest/case.py", line 831, in assertEqual
    assertion_func(first, second, msg=msg)
  File "/usr/lib64/python3.9/unittest/case.py", line 1037, in assertListEqual
    self.assertSequenceEqual(list1, list2, msg, seq_type=list)
  File "/usr/lib64/python3.9/unittest/case.py", line 1019, in assertSequenceEqual
    self.fail(msg)
  File "/usr/lib64/python3.9/unittest/case.py", line 670, in fail
    raise self.failureException(msg)

AssertionError: Lists differ: ['-rwxr-xr-x 6000 6000 /etc/init.d/mdmonitor', '-rw-r-[10124 chars]n.8'] != ['-rwxr-xr-x root root /etc/init.d/mdmonitor', '-rw-r-[10124 chars]n.8']

First differing element 0:
'-rwxr-xr-x 6000 6000 /etc/init.d/mdmonitor'
'-rwxr-xr-x root root /etc/init.d/mdmonitor'

This is due to a version of pseudo without statx being used on a system where
ls uses statx, hence the files are displayed as 6000.6000 instead of root.root.

Avoid this by always building pseudo-native for the specific distro in question rather
than using a universal sstate feed.

This hopefully fixes one of the mysterious AB-INT issues.

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 6e3785a3f1f3cf68f5fe101cd6bebe91db165973)
Signed-off-by: Steve Sakoman <steve@sakoman.com>
meta/classes/base.bbclass
meta/classes/sstate.bbclass
meta/lib/oe/sstatesig.py