Benny Halevy

One example for apparent nested Segfault is: ``` (gdb) thread 2 [Switching to thread 2 (Thread 0x7f10cabff640 (LWP 2656476))] #0 0x00007f10cd36d498 in uw_frame_state_for () from /lib64/libgcc_s.so.1 (gdb) bt #0 0x00007f10cd36d498...

The server can try to send an error back, and if that failed too then maybe closing the connection would be better.

Fixes #1148 It doesn't cause any scylla issue, but rather is triggered by https://github.com/scylladb/scylla/pull/11066/commits/0428b34215e6851db78f11d3c5b8bcc9610c8051

> And for the record: I can only produce these stalls with ASAN turned on -> probably malloc delays causing it... @elcallio any progress? This is killing us in CI,...

> @bhalevy - the linked issue is a coredumpo in scylla. How is that related to this test? I mentioned this issue in https://github.com/scylladb/scylladb/issues/11252 just because of the general frustration...

I see `on_internal_error_noexcept()` similar to `assert()` that may return under certain conditions (`NDEBUG`). Maybe adding `abort_on_internal_error` makes sense since there are few places already we call `on_internal_error_noexcept` immediately followed by...

In v2 (56c733e1): added const and noexcept qualifiers

The plan is to replace ```c++ bool compaction_manager::can_perform_regular_compaction(replica::table* t) { if (!can_proceed(t) || t->is_auto_compaction_disabled_by_user()) { return false; } auto& lk = get_compaction_state(t).lock; auto could_lock = lk.try_read_lock(); if (!could_lock) { return...

> > ``` > > auto& lk = get_compaction_state(t).lock; > > return lk.can_read_lock(); > > ``` > > OK, so the goal is to find out it the lk can...

> > > > ``` > > > > auto& lk = get_compaction_state(t).lock; > > > > return lk.can_read_lock(); > > > > ``` > > > > > >...