Segfaults

[simendsjo]

I seem to get segfaults at various locations again when remote connections are hanging up. Haven't tried a minimal example, but just holding F5 pressed in a browser does the trick for me.

libevent2 -> lev_unlock_mutex -> mtx.mutex.unlock

tried adding some asserts, and then I got some errors here

Program received signal SIGSEGV, Segmentation fault.
0x0000000000715000 in rt.deh2.terminate() ()

[mihails-strasuns]

Please always mention vibe.d and dmd version in question - I have been recently tracking down some stuff and feels like that any combo of [last dmd release, dmd master] [last vibe.d release, vibe.d master] can produce its own quirks.

[simendsjo]

vibe.d master, dmd 2.063.2, linux x64.

I tested a bit here, and it might seem like some corruption bug of some sort.
Adding assert(mtx !is null) after casting from lock succeeds. adding assert(mtx.mutex !is null) in the else block in question throws.
But if I change the else to else if(mtx.mutex !is null) it gives segfault at that line.

[s-ludwig]

Fixed by d041fd4 as far as I could reproduce it. It's strange that didn't occur more often as it must have been there quite from the beginning and was quite easy to trigger now.

[simendsjo]

I'll test this later. BTW, vibe has always triggered segfaults here - think I even reported a very similar report this spring.

[s-ludwig]

Yeah, it may be the same as the concurrent requests/mysql-native bug, but I could never reproduce that one. But now it failed instantly with the help of a little sleep() - something that didn't help back then.

[s-ludwig]

Closing for now.