LTP-Realtime hangs on Arndale when the full suite is run

Working sub lists are now handled by a yaml script that explicity lists what shoudl be run.

Assigned resouses to look at this https://cards.linaro.org/secure/RapidBoard.jspa?rapidView=131&view=detail&selectedIssue=LNG-525

Testing is in progress to discover and resolve the reasons why some LTP-realtime tests are 'hanging' on Arndale. This effort is also focused on expanding the number of LTP realtime tests which LNG runs to encompass the full suite of available tests.

Some priority inheritance tests are hanging because conditional compilation prevents complete initialization of PI mutexes used in the tests. This issue is being addressed and additional issues will be likewise resolved as they are discovered.

Conditional compilation issues were resolved locally... now hanging in func/prio-wake test... apparently some issue with condition variables deep in the futex library... investigating this. Everything else seems at least to run to completion. Will submit patches as soon as prio-wake issues are resolved.

Still working on isolation of prio-wake issues. Intend to check x86_64 to see if this fails there. If it fails on both ARM and x86_64 then it's probably a code issue in the NPTL / futex code. If x86_64 is okay but ARM fails then it's likely an issue with the assembly code generated by the ARM compiler.

prio-wake functions properly on x86_64. GDB tracing inside a LAVA hacking session appears to show a race condition in the mutex locking code between multiple waking waiters on armv7a. Submitted a separate bug on this issue via BUGZILLA.

All the other ltp-realtime tests run to completion on linux-lng-preempt-rt-v3.14.x - see results bundle:
https://validation.linaro.org/dashboard/permalink/bundle/5df34d85858a91fdd7ff4f617b3f6388b4f205c8/

I made a number of patches to ltp-realtime via the bitbake recipe in 'jenkins-setup/openembedded-core/meta/recipes-extended/ltp'. These patches are appropriate to the OE environment but maybe not so much in the more generic LTP environment - since LTP normally expects to be configured and built from a tarball on the same system where it will be executed, and this is not the situation in the OE build environment.

Accordingly I will submit my recipe patches to the OE maintainers, but unless directed otherwise I am disinclined to bother the LTP maintainers with them.

Patches have been submitted upstream to both LTP and OE-core maintainers, and are currently undergoing review and discussion pending hopeful approval.

Should mention the issue to Fathi as it as some of the issue is related to our build process

migrated to https://bugs.linaro.org/show_bug.cgi?id=339