On Wednesday, 1 April 2020 at 02:08:09 UTC, mitchell wrote:
Hello,
I've just finished migrating a project from GDC/Makefile to LDC
with dub. The program now compiles and runs, and works fine
until such time as the following error occurs:
Aborting from core/sync/mutex.d(147) Error: pthread_mutex_init
failed.
out of GDB, the whole text is:
Aborting from core/sync/mutex.d(147) Error: pthread_mutex_init
failed.
Thread 1 "squares" received signal SIGABRT, Aborted.
[Switching to Thread 0x7ffff74a7580 (LWP 7576)]
__GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:50
50 ../sysdeps/unix/sysv/linux/raise.c: No such file or
directory.
and the backtrace is (sorry for info dump):
#0 __GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:50
#1 0x00007ffff76eb535 in __GI_abort () at abort.c:79
#2 0x00007ffff7ac4fa7 in
_D4core8internal5abortQgFNbNiNfMAyaMQemZv ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#3 0x00007ffff7acd7e3 in core.sync.mutex.Mutex.~this() ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#4 0x00007ffff7af12ae in rt_finalize2 () from
/lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#5 0x00007ffff7adb7bf in
_D2gc4impl12conservativeQw3Gcx5sweepMFNbZm ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#6 0x00007ffff7ad89b2 in
_D2gc4impl12conservativeQw3Gcx11fullcollectMFNbbZm ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#7 0x00007ffff7ada126 in
_D2gc4impl12conservativeQw3Gcx10smallAllocMFNbhKmkZPv ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#8 0x00007ffff7ad6c44 in
_D2gc4impl12conservativeQw14ConservativeGC__T9runLockedS_DQCeQCeQCcQCnQBs12mallocNoSyncMFNbmkKmxC8TypeInfoZPvS_DQEgQEgQEeQEp10mallocTimelS_DQFiQFiQFgQFr10numMallocslTmTkTmTxQCzZQFcMFNbKmKkKmKxQDsZQDl () from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#9 0x00007ffff7ad92df in
_DThn16_2gc4impl12conservativeQw14ConservativeGC6qallocMFNbmkxC8TypeInfoZS4core6memory8BlkInfo_ () from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#10 0x00007ffff7addbd0 in gc_qalloc () from
/lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#11 0x00007ffff7ac9fb2 in
_D4core6memory2GC6qallocFNaNbmkxC8TypeInfoZSQBqQBo8BlkInfo_ ()
from /lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#12 0x00007ffff7af0519 in _d_newarrayU () from
/lib/x86_64-linux-gnu/libdruntime-ldc-shared.so.82
#13 0x00005555556b9682 in
_D7squares4core5coord5Coord9adjacentsMxFNdZASQBrQBmQBkQBh
(this=...)
at
/home/mitchell/Apps/D/games/squares/src/squares/core/coord.d:102
There are several frames above #13 here, but #13 is the last
line of "my" code, and the deeper ones (appear?) to be runtime
and phobos stuff.The lines immediately around coord.d:102 are:
Coord[] adjacents() const @property
{
return [ right, up, left, down ]; // line 102
}
which are defined in a struct called "Coord". I suspect that
there isn't something wrong specifically with this line, and
instead something more "amazingly" wrong. Given that this
worked fine with the previous (and older) version of GDC which
I used before my migration, I don't know what could be wrong.
I'm just not entirely certain where to start looking for
problems. If this issue is over-broad, I'm sorry. Elsewhere in
the program, I make use of the core.thread and Thread things,
however that implementation seems fairly straightforward and
fine, and never had issues under the old compiler.
Any and all help is appreciated!
I have (surprisingly) figured out the issue myself. I guess I
didn't think about this before, but the background thread
(elsewhere in the application) keeps a few mutexes such that it
can pass work to the main thread (for changing UI things, and
such) and it appears that after the background task is completed,
there is at least one mutex which was still held in a locked
state. The error arises (as can been seen from the stack trace)
when that mutex is being destructed. As my underlying system is
unix, the pthread object beneath mutex is destroyed, but
according to the documentation
(https://linux.die.net/man/3/pthread_mutex_destroy) destroying a
mutex while it's still locked can cause an error. After I fixed
my implementation to ensure that all mutexes are unlocked on the
appropriate threads before they can be destructed, the issue went
away. Apparently my old compiler either handled that issue for me
automatically, or it just wasn't being detected.
TLDR: pebkac