summaryrefslogtreecommitdiffstats
path: root/bitbake/lib/bb/tests/fetch.py
diff options
context:
space:
mode:
authorRichard Purdie <richard.purdie@linuxfoundation.org>2022-03-25 14:51:43 +0000
committerRichard Purdie <richard.purdie@linuxfoundation.org>2022-03-26 09:27:43 +0000
commit7e273d09d0fe30c3254edd864fb822777109cddc (patch)
tree26bbdfd982c70f7a1ab978a83cf4744132545c6a /bitbake/lib/bb/tests/fetch.py
parentcff6c1a18d56e20d99d2331704dd490940c386e3 (diff)
downloadpoky-7e273d09d0fe30c3254edd864fb822777109cddc.tar.gz
bitbake: cooker: Fix inotify watches causing memory resident bitbake corruption
Thanks to great debugging from pavel@zhukoff.net we had a simpler reproducer for the corruption see in oe-selftest when using BB_SERVER_TIMEOUT=60, i.e. with bitbake in memory resident mode. This was effectively: oe-selftest -r devtool.DevtoolUpgradeTests.test_devtool_upgrade devtool.DevtoolUpgradeTests.test_devtool_upgrade_git -j 1 -K The issue is that if directories are removed (such as workspace), if they are added again, we don't have the watches in place any more. This patch adds some slightly paranoid checks to ensure we do the correct things for directory additions and removals (we track directories, not files specifically to avoid running out of watches). [YOCTO #14023] (Bitbake rev: 2c414f659d793d732041614caedd773959eb4f27) Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Diffstat (limited to 'bitbake/lib/bb/tests/fetch.py')
0 files changed, 0 insertions, 0 deletions