This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
RE: Question: Is Dwarf CFI support required for new target?
- From: "Breazeal, Don" <Don_Breazeal at mentor dot com>
- To: Simon Marchi <simon dot marchi at polymtl dot ca>
- Cc: "gdb-patches at sourceware dot org" <gdb-patches at sourceware dot org>
- Date: Fri, 1 Jun 2018 22:57:17 +0000
- Subject: RE: Question: Is Dwarf CFI support required for new target?
- References: <0410836b-5744-f823-80c5-057c8f113914@codesourcery.com> <226ee236120250e03a0539f96eca720d@polymtl.ca>
> -----Original Message-----
> From: Simon Marchi [mailto:simon.marchi@polymtl.ca]
> Sent: Wednesday, May 30, 2018 11:49 AM
> To: Breazeal, Don <Don_Breazeal@mentor.com>
> Cc: gdb-patches@sourceware.org
> Subject: Re: Question: Is Dwarf CFI support required for new target?
>
> On 2018-05-29 12:14, Don Breazeal wrote:
> > Hi,
> > I have a question about the requirements for GDB port to a new target:
> > is support for Dwarf Call Frame Information required for acceptance,
> > or is prologue analysis sufficient?
> > Thanks
> > --Don
>
> Hi Don,
>
> Can you provide more context? I can interpret your question in two
> ways:
>
> - Your producer does not produce CFI, so you would be forced to always
> use prologue analysis for unwinding. In that case, I guess it's fine,
> GDB should do its best with the information at its disposal.
> - Your producer produces CFI but GDB for your architecture won't use it.
> In that case, I would ask why don't you use it. It's generally more
> reliable and less hair-pulling than prologue analysis.
>
> Simon
Hi Simon,
My scenario is currently a little less concrete than those you describe, but this feedback is helpful.
Thanks
--Don