Crash with SIGPIPE while loading previous messages

[DataBeaver]

I noticed that when scrolling upwards through a long channel history, Mirage crashes pretty quickly. There's no messages in the terminal but gdb reveals it to be a SIGPIPE:

Thread 7 "QPythonWorker" received signal SIGPIPE, Broken pipe.
[Switching to Thread 0x7fffd562a700 (LWP 13043)]
__GI___libc_write (nbytes=24, buf=0x7fffddec9c10, fd=41) at ../sysdeps/unix/sysv/linux/write.c:26
26      ../sysdeps/unix/sysv/linux/write.c: No such file or directory.
(gdb) bt
#0  __GI___libc_write (nbytes=24, buf=0x7fffddec9c10, fd=41) at ../sysdeps/unix/sysv/linux/write.c:26
#1  __GI___libc_write (fd=fd@entry=41, buf=buf@entry=0x7fffddec9c10, nbytes=nbytes@entry=24)
    at ../sysdeps/unix/sysv/linux/write.c:24
#2  0x00007fffd6c10620 in __pyx_f_6uvloop_4loop_8UVStream__try_write (__pyx_v_self=0x7fffcc0bb860, __pyx_v_data=<optimized out>)
    at uvloop/loop.c:87771
#3  0x00007fffd6c21d62 in __pyx_f_6uvloop_4loop_8UVStream__exec_write (__pyx_v_self=__pyx_v_self@entry=0x7fffcc0bb860)
    at uvloop/loop.c:88613
#4  0x00007fffd6c24e08 in __pyx_f_6uvloop_4loop_4Loop__exec_queued_writes (__pyx_v_self=0x7fffe0625820) at uvloop/loop.c:15829
#5  0x00007fffd6c50360 in __pyx_f_6uvloop_4loop_6Handle__run (__pyx_v_self=__pyx_v_self@entry=0x7fffd6f02670)
    at uvloop/loop.c:56458
#6  0x00007fffd6c53374 in __pyx_f_6uvloop_4loop_cb_check_callback (__pyx_v_handle=<optimized out>) at uvloop/loop.c:76086
#7  0x00007fffd6c62d61 in uv__run_check (loop=loop@entry=0x7fffe0234e40) at src/unix/loop-watcher.c:67
#8  0x00007fffd6c606dd in uv_run (loop=0x7fffe0234e40, mode=mode@entry=UV_RUN_DEFAULT) at src/unix/core.c:376
#9  0x00007fffd6c1c100 in __pyx_f_6uvloop_4loop_4Loop___run (__pyx_v_self=0x7fffe0625820, __pyx_v_mode=UV_RUN_DEFAULT)
    at uvloop/loop.c:13914
#10 0x00007fffd6c254c2 in __pyx_f_6uvloop_4loop_4Loop__run (__pyx_v_self=0x7fffe0625820, __pyx_v_mode=UV_RUN_DEFAULT)
    at uvloop/loop.c:14299
#11 0x00007fffd6b8c315 in __pyx_pf_6uvloop_4loop_4Loop_24run_forever (__pyx_v_self=0x7fffe0625820) at uvloop/loop.c:27024
#12 __pyx_pw_6uvloop_4loop_4Loop_25run_forever (__pyx_v_self=0x7fffe0625820, unused=<optimized out>) at uvloop/loop.c:26827
#13 0x00007fffebc79297 in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#14 0x00007fffeba51ffd in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#15 0x00007fffeba53274 in _PyEval_EvalFrameDefault () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#16 0x00007fffeba5d0eb in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#17 0x00007fffebc80bab in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#18 0x00007fffebc805cc in PyVectorcall_Call () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#19 0x00007fffeba57b29 in _PyEval_EvalFrameDefault () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#20 0x00007fffeba5d0eb in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#21 0x00007fffeba51ffd in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#22 0x00007fffeba53274 in _PyEval_EvalFrameDefault () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#23 0x00007fffeba5d0eb in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#24 0x00007fffeba51ffd in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#25 0x00007fffeba53274 in _PyEval_EvalFrameDefault () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#26 0x00007fffeba5d0eb in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#27 0x00007fffebc80bab in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#28 0x00007fffebc805cc in PyVectorcall_Call () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#29 0x00007fffebaf23dd in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#30 0x00007fffebb57d67 in ?? () from /usr/lib/x86_64-linux-gnu/libpython3.8.so.1.0
#31 0x00007ffff5c16f27 in start_thread (arg=<optimized out>) at pthread_create.c:479
#32 0x00007ffff60442ef in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

I'm not sure if this is a bug in uvloop, the nio library or Mirage itself. Disabling uvloop fixes it, as does adding signal(SIGPIPE, SIG_IGN); at the top of main().

[mirukana]

I think I'll drop uvloop for the next release, it's already giving me headaches with impossible to catch socket.gaierror exceptions when network issues occur and I don't have the skills or will to debug its code.

[mirukana]

0.5.0 has been released, is the crash gone out of the box?

[DataBeaver]

Apologies for the delay, I've had a busy few days. Version 0.5.0 does not exhibit the SIGPIPE crash.