[patches] Re: dwarf2/ACCUMULATE_OUTGOING_ARGS fix

[Jason Merrill]

>>>>> "Jan" == Jan Hubicka <jh@suse.cz> writes:

>> >>>>> "Jan" == Jan Hubicka <jh@suse.cz> writes:
>> 
>> >> > For use in debugger and in unwinding library for x86_64.  For instance
>> >> > the garbage collector needs to unwind stack at any point.
>> >> 
>> >> It does?  Why?  How?  What is this unwinding library of which you speak?
>> 
>> > When you have non-conservative garbage collector over multithreaded
>> > application.  When such beast is activated it needs to parse stacks of
>> > all threads and thus needs to unwind stack at any state of execution.
>> 
>> Ah.  I suppose the additional detail could be controlled with a flag (as it
>> already is somewhat, based on -fasynchronous-exceptions).

> I am having this handled in x86_64 specific way - I've added
> ASYNCHRONOUS_DWARF2_UNWIND_INFO, that enables this behaviour and have
> -mno-unwind-info to disable it (for instance for kernel compilation).

I really don't think this extra detail should be emitted by default.  How
many programs do you know that use both a non-conservative garbage
collector and threads?

I also don't think it should be target-specific.

> If you have idea about better interface, I may change it.

One possibility would be to emit the full information in .debug_frame, and
require the GC unwinder to look it up using libelf (as SGI do for EH
unwinding, too).

Jason