This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PING][PATCH] generate DWARF2 unwind information for epilogues
- From: Andrew Haley <aph at redhat dot com>
- To: Nathan Froyd <froydnj at cs dot rice dot edu>
- Cc: gcc-patches at gcc dot gnu dot org, johnmc at cs dot rice dot edu
- Date: Wed, 8 Mar 2006 10:02:30 +0000
- Subject: Re: [PING][PATCH] generate DWARF2 unwind information for epilogues
- References: <20060307203753.GA3973@blunux.cs.rice.edu>
Nathan Froyd writes:
> Please see:
>
> http://gcc.gnu.org/ml/gcc-patches/2006-02/msg01091.html
>
> It's not mentioned in the original email, but it fixes:
>
> * Bug 12990 "CFA not tracked in epilogues"
> * Bug 18749 "GCC fails to mark stack-popping instruction in unwind-info"
A few observations.
Firstly, a 1.7% increase in the total size of binaries is quite a lot.
It indicates a much greater than 1.7% bloating of the unwinder
information.
Secondly, you say that this was bootstrapped on x86-64, but there also
seems to be a lot of 32-bit x86 code.
Finally, this means that the x86/x86_64 unwinder generates unwinder
data for epilogues but the other targets don't. It seems to me that
the decision whether this data should be generated needs to be made by
gcc maintainers as a whole, and we should try to be consistent across
targets.
Andrew.