[Lldb-commits] [PATCH] D58727: https://reviews.llvm.org/D58394 done right

Jim Ingham via lldb-commits lldb-commits at lists.llvm.org
Mon Mar 4 17:26:37 PST 2019

The event we are stopping for is the stop to single step over a breakpoint.  We should have decided not to broadcast that to the public event queue, since it's an implementation stop of the "continue".  It almost looks like the stop hook is getting triggered by a not-broadcast event, though that doesn't make sense.

I'll have to look at this more closely, I'm in the middle of something else right now.  Thanks for adding the Unsupported...


> On Mar 2, 2019, at 11:54 AM, Pavel Labath <pavel at labath.sk> wrote:
> On 01/03/2019 14:55, Pavel Labath via Phabricator wrote:
>> labath accepted this revision.
>> labath added a comment.
>> This revision is now accepted and ready to land.
>> Yes, that did the trick. The new test is passing on linux now as well.
>> Regarding BTW, I see the option to "reopen a revision" in the drop down menu at the bottom of the page (the same one you use to accept the revision and stuff). AFAICT, even I am able to reopen that other revision.
> Hmm.. it seems this did not fully fix the problem (at least on linux), because I am still seeing random failures (usually when the system is under load).
> I've captured the logs of both failing and successful runs, but I'm not exactly sure what to make of it. Things start to get interesting around line 22610 in bad.log, where we do a "vCont;s" to step over the breakpoint. However, it seems that the public thread does not wait for the single step to complete, and immediately starts to compute backtraces of threads (as if we just performed the final stop). Around lines 22700--22800 you can see that the public thread's backtrace computing runs concurrently with the ShouldStop processing on the private thread, which doesn't seem right to me. Things get very confusing after that, but the end result seems to be that the process gets killed (maybe because we've reached the end of the batch file) without it having completed all prescribed stops.
> Do you have any idea what's going on here?
> PS: For the time being, I've reinstated the "UNSUPPORTED: linux" stanza on this test as the test fails pretty frequently for me.
> regards,
> pavel
> <bad.log.bz2><good.log.bz2>

More information about the lldb-commits mailing list