scheduler: Workaround negative nsecs bug in boost's wait_until #18271

pull luke-jr wants to merge 1 commits into bitcoin:master from luke-jr:wrkarnd_boost_wait_until changing 1 files +14 −2

luke-jr commented at 10:43 PM on March 5, 2020: member

Some boost versions have a bug that can cause a time prior to system boot (or wake from sleep) to throw an exception instead of return timeout See https://github.com/boostorg/thread/issues/308

Fixes #18227

Cleanly merges to 0.14+
DrahtBot commented at 1:42 AM on March 6, 2020: member



The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.



Conflicts

No conflicts as of last run.
ajtowns commented at 2:46 AM on March 6, 2020: member

I think travis is failing because this catches boost::thread_interrupted and sometimes doesn't rethrow it. Checking timeToWaitFor <= ::now() prior to calling wait_until might work better?

scheduler: Workaround negative nsecs bug in boost's wait_until

Some boost versions have a bug that can cause a time prior to system boot (or wake from sleep) to throw an exception instead of return timeout
See https://github.com/boostorg/thread/issues/308

ed0223ec59

luke-jr force-pushed on Mar 6, 2020
luke-jr commented at 3:55 AM on March 6, 2020: member

That could break in a race (suspend after check, but before wait_until).

Fixed it by always rethrowing boost::thread_interrupted reliably.
MarcoFalke commented at 7:47 PM on March 6, 2020: member

Please enable the mockscheduler unit test again
luke-jr closed this on Mar 6, 2020
DrahtBot locked this on Feb 15, 2022

Contributors

Linked (view graph)

#18227 scheduler test causing pthread_cond_timedwait: Invalid argument error

scheduler: Workaround negative nsecs bug in boost's wait_until #18271

Conflicts