Exception size reduction patch

H.J. Lu hjl@lucon.org
Fri Apr 24 10:11:00 GMT 1998


> 
> I think it would make more sense to use __attribute__ ((weak)), which
> should work for any target that supports weak declarations, and which
> will be harmless if it doesn't.  This is better than checking for
> HANDLE_SYSV_PRAGMA, since not every target that supports weak declarations
> supports HANDLE_SYSV_PRAGMA.  I don't think you need to conditionalize the
> checks for non-zero function addresses.

Will do.

> 
> It isn't clear to me why you moved all of the eh.o stuff from libgcc2.c to
> frame.c.  That just seems to make the patch larger than it needs to be.
> 
> 	If we want, we can separate sjlj exception from frame-based exception
> 	so that -fsjlj-exceptions won't include the frame info.
> 
> It isn't clear why this is a concern.  Did you perhaps increase sjlj executable
> size by putting the eh.o stuff in frame.c?  And now you are suggesting a way
> to avoid this?
> 
> Or are you saying that this patch doesn't decrease sjlj executable size, and
> that further changes do?  If that is the case, then I would suggest making one
> change at a time.
> 

Ok. I will create a new file, libgcc2.h, which has the definitions
and declarations used by both libgcc2.c and frame.c.

I have one quesion. I didn't find find_exception_handler () was
ever used in the sjlj exception. Should I move it into the DWARF2
unwind exception?

Thanks.


-- 
H.J. Lu (hjl@gnu.org)



More information about the Gcc mailing list