PING^2: [RFA/RFC] fix PR gdb/22670 (pb looking up some symbols when they have a linkage name)
Joel Brobecker
brobecker@adacore.com
Tue Mar 27 14:02:00 GMT 2018
> >>> I ended up choosing this approach. lookup names is still a bit
> >>> foggy for me, so the comments I wrote and/or the name of the
> >>> new language_defn attribute might be a bit off. As always, I am
> >>> grateful for suggestions :).
>
> I looked again at this today, and I couldn't think of a better
> option right now. So I think you should go ahead and push this in.
Thanks, Pedro.
I re-tested the patch on x86_64-linux, and then pushed it to master.
> Note that this C++ mangled form like in "symada__cS", is not
> actually how C++ symbols are mangled nowadays. The modern Itanium
> mangling scheme has linkage names that always start with "_Z". In this
> case, "void symada(char, signed)" mangles as "_Z6symadaci".
Interesting.
> It's possible that we no longer need to support that older (and
> ambiguous with Ada and other languages) mangling scheme for C++, I
> don't know. I wish Ada linkage names also had some kind of leading
> prefix, but that can't happen without an ABI break, of course...
Yeah, it's a good idea, but making that kind of change happen is
going to be tough, especially since we know there are consumers
of our debugging information that we have no control of. I'll mention
it to the compiler team, though.
Thanks Pedro.
--
Joel
More information about the Gdb-patches
mailing list