This is the mail archive of the
gdb-patches@sourceware.cygnus.com
mailing list for the GDB project.
Re: [RFA] SH prologue recognition update
- To: Andrew Cagney <ac131313 at cygnus dot com>
- Subject: Re: [RFA] SH prologue recognition update
- From: Chris Faylor <cgf at cygnus dot com>
- Date: Fri, 21 Apr 2000 10:29:40 -0400
- Cc: gdb-patches at sourceware dot cygnus dot com
- References: <20000420231327.A4674@cygnus.com> <38FFEE7F.7628FA60@cygnus.com>
On Fri, Apr 21, 2000 at 04:00:31PM +1000, Andrew Cagney wrote:
>Chris Faylor wrote:
>>
>> The following changes were necessary to get sh-tdep.c working with the
>> Windows CE SH target. They seem to accomodate the prologue found in
>> more recent version of gdb. I've also added an additional defensive
>> check to sh_find_callers_reg.
>
>Is there a need for the (dreaded?) heuristic_fence_post in this code?
>What is the chance of GDB just plowing through memory looking for the
>end of the prologue?
It's already bounded. It should only look for 28 instructions or so.
This seems to be slightly more than the longest prologue I could produce
but it's probably not perfect.
>The other question is, does WinCE/SH GCC output the same prologe as for
>other SH targets such as sh-coff(?) and sh-elf? Playing around with the
>prologue analysis code is hairy so I'm trying to figure out how much
>testing is practical.
AFAIK, yes. I actually imported the latest sh/gcc code into the WinCE branch.
It looks pretty certain that the prologue may have changed and that my
code makes at least a slightly better stab at finding it.
However, once again (hangs head in shame), I could have run this through the
test suite but I didn't. I'll try that now.
cgf