This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [RFC] fullname attribute for GDB/MI stack frames
On Thu, May 05, 2005 at 12:45:43PM -0400, Bob Rossi wrote:
> O, I see. I was going to add the regex and update mi-file.exp. However,
> I noticed that mi2-file.exp also had a fullname field and realized I
> didn't know if the mi2 files were going to be modified.
Oh, those testcases. Go ahead.
> I'm looking now, and just realized that Andrew didn't let me modify
> mi2-file.exp to add -file-list-exec-source-files even though it was
> compatible with the MI2 protocol. He only wanted me to add it to
> mi-file.exp. So it looks like your initial idea was the philosophy
> Andrew was taking, which is: MI2 is stable and it should not be modified.
> I am going to take that approach with this patch.
The one doesn't follow from the other. -file-list-exec-source-files
shouldn't be tested as a part of MI2, because MI2 did not include it at
its release.
However, I don't see a problem with allowing extra fields, assuming
that you are right about frontends being accepting of this. It isn't
clearly spelled out in the manual, one way or another.
> I really wish the MI release philosophy could be spelled out some where.
> It seems as if Andrew considered MI versions to not be at the protocol
> level. For instance, MI2 and MI3 could both use the same protocol, except
> that MI3 would have more functions and fields than MI2. Does this sound
> correct to you? This is part of the confusion I had when trying to figure
> out how GDB would handle multiple MI protocols.
I'm not the one you should be asking since I don't think all the
multiple protocols work is/was useful. The model just has too many
question marks in it when you consider all those busy beavers
developing on the _rest_ of GDB. You're lucky if you get even one MI
protocol to work.
--
Daniel Jacobowitz
CodeSourcery, LLC