labath added a comment.

What lldb does not like the most is a stop info with a signal "0" -- that 
causes a resume. Having no stop info whatsoever seems to work just fine -- and 
this is already what happens when there is no exception record around. I was 
surprised by that distinction, but we already have tests with minidumps like 
those, and they seem to work ok, so I thought it'd make sense to do the same 
for the case where we have an exception record, but it contains no signal (as 
opposed to e.g. inventing a SIGSTOP out of thin air). We already have a test 
for that in this patch.

Comment at: lldb/source/Plugins/Process/minidump/ProcessMinidump.cpp:264
     stop_info = StopInfo::CreateStopReasonWithSignal(
+        *stop_thread, signo);
clayborg wrote:
> Are exceptions on linux always just a signal number? Mac can be a mach 
> exception or a signal. Might be worth looking at the breakpad sources to see 
> if they ever fill in anything else for linux other than a signal. 
> Also, there is more info in the exception record that needs to be passed 
> along. Like for a SIGSEGV, the address should be in the exception info 
> somewhere.
Having the address (if it is there) would be nice, but I think that's a 
separate bug/feature, as this patch is about what happens when there is no 
exception record around.

  rG LLVM Github Monorepo


lldb-commits mailing list

Reply via email to