linux/poky.git/meta/lib, branch uninative-2.8

abi_version/sstatesig: Introduce HASHEQUIV_HASH_VERSION

2020-02-28T23:11:28+00:00

We've found we need a way to cause a change in signatures and move to a new hash 'namespace' with hashequiv. This introduces a variable which allows us to do this. (From OE-Core rev: 1c2d5aeae2fa2dfc16bc60b6cf0b310f826337fd) Signed-off-by: Richard Purdie

ptest: add an option to enforce a failure (e.g. ptests are required to pass)

2020-02-25T10:41:22+00:00

This can be useful with a more specific, targeted and robust set of ptest packages; the benefit is that ptest regressions are caught as they happen and can be more easily traced to changes that caused them. The existing AB ptest image continues to be expected to fail, my observation of the AB runs is that the full set of ptests is not robust enough (particularly around socket/network related tests) and sporadically fails in random places. This can probably be addressed by making ptests exclusive to a worker (e.g. there is no other workload happening at the same time as ptests). (From OE-Core rev: 102afcc7ea7d7673183f272a29ab233084ad168b) Signed-off-by: Alexander Kanavin Signed-off-by: Richard Purdie

oeqa/logparser: correctly set test result markers

2020-02-22T23:57:27+00:00

I spotted this by noticing ptests started to unexpectedly pass, which shouldn't happen yet. (From OE-Core rev: 1cd6e9a8fdfef927916c6cea2371fc0430b1ffdf) Signed-off-by: Alexander Kanavin Signed-off-by: Richard Purdie

oeqa/logparser: Fix performance issues with ptest log parsing

2020-02-21T09:39:00+00:00

On the autobuilder a ptest log with 2.1 million lines took around 18 hours to process. This is clearly crazy. We can tweak the processing code to: a) Stop repeatedly joining large strings together (append to a list instead) b) Use one startswith expression instead of multiple re.search() operations With this change it takes 18 hours down to around 12s. [YOCTO #13696] (From OE-Core rev: 01c9f40b979e36a53e789a6bedd89b6d9557dce3) Signed-off-by: Richard Purdie

oeqa/selftest: Drop 'backup' code and SIGTERM handler

2020-02-21T09:39:00+00:00

Now selftest is using its own copied build directory, we can stop worrying about copying files around as backup, and drop the SIGTERM handler to try and restore them, simplifying the code. (From OE-Core rev: b8ea8a910267fee4bb9e57f24ba829064e22d016) Signed-off-by: Richard Purdie

oeqa/selftest: Standardise seperate builddir for concurrent and non-concurrent selftest

2020-02-21T09:39:00+00:00

Currently oe-selftest reuses the current build directory and the concurrent version run with -j does not. Standardise and use a separate new build directory in both cases. This will lead to simpler code and more reliable user run tests. (From OE-Core rev: 50238b0717b04e0a1fa69d618e8c8aa8445a80b0) Signed-off-by: Richard Purdie

selftest/context: Avoid tracebacks from tests using multiprocessing

2020-02-21T09:39:00+00:00

We can see tracebacks where the SIGTERM handler catches things it shouldn't. Avoid exit(1) unless we're the process that it was intended for. [YOCTO #13664] (From OE-Core rev: d9c62ffac611310efd47ed6397d31dccb72fe868) Signed-off-by: Richard Purdie

oescripts.py: use unittest.SkipUnless for class skip

2020-02-20T13:00:28+00:00

Yocto Bug: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13663 Prior to this fix, the try/except block in OEScriptTests' setUpClass did both the import and the raise unittest.SkipTest for the subsequent OEPybootchartguyTests. The consequence of the this meant that, when cairo was not available on the test host, the raise was counted only once as a class setup issue, instead of once for each of the tests to be run, resulting in incorrect skip counts in the test results. Instead of raising unittest.SkipTest in setUpClass, the more proper method is to use the unittest.skipUnless method along with importlib.util.find_spec to check for the presence of the python3-cairo module. This allows the entire class to be skipped if cairo is not present, while also correctly passing along the skip state for each of the class's tests, and avoiding the UNKNOWN results to appear in cases where the -j option was provided to oe-selftest, but the tests were still skipped. (From OE-Core rev: 10c2216cb6fb61c74babd8bd17cb3481754b7475) Signed-off-by: Trevor Gamblin Signed-off-by: Ross Burton Signed-off-by: Richard Purdie

oeqa: reproducible: Include jquery in results

2020-02-19T21:28:06+00:00

Includes jquery in the output directory to make the diffoscope output easier to navigate (From OE-Core rev: ff7c386455848e497ef40f9fee2be87d5171e203) Signed-off-by: Joshua Watt Signed-off-by: Richard Purdie

concurrencytest.py: add outSideTestaddSkip for subunit

2020-02-18T23:53:55+00:00

see: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13663 When running oe-selftest in concurrency mode (e.g. with oe-selftest --run-tests oescripts.OEPybootchartguyTests -j 4), if a skip occurred during setUpClass() rather than within individual tests, the entire suite would show "UNKNOWN" as each test's result. This is because subunit doesn't know how to handle skips outside of individual tests. An example of where this occurs is when running the above call to oe-selftest in concurrency mode on a host machine that does not have python3-cairo installed. Patch subunit inside concurrencytest.py to provide a method called outSideTestaddSkip, which will allow subunit to correctly detect the skip in setUpClass(). (From OE-Core rev: 9b8734b584d6e8d9c32ff2a721b29f3f3e61cca7) Signed-off-by: Trevor Gamblin Signed-off-by: Richard Purdie