Re: [PATCH]: New maintainer command "flushregs"

[Steven Johnson <sbjohnson at ozemail dot com dot au>]

Michael Snyder wrote:
> 
> I added this to help with testing register fetch and store.
> 
> 2000-09-01  Michael Snyder  <msnyder@cleaver.cygnus.com>
> 
>         * regcache.c (reg_flush_command): New function.  Maintainer-mode
>         command, flushes GDB's register cache, for testing purposes.
> 
[SNIP PATCH]

This is needed for more than testing purposes. When doing embedded
debugging, there are times when the registers of the target can change
but GDB doesn't know it. At those times GDB needs to be "Forced" to
flush it's caching. This command is good. Should it be maintenance or
under some other category?

To Illistrate when GDB can get out of sync with a Target:

GDB Has no <standard> command to force a target to reset. (As in
Hardware Reset).

I Implemented this in the stub and use a Stub Command from GDB to
perform it.

The problem is, after a hardware Reset, all the registers have changed
but GDB does not know it. GDB Never looses connection to the target
because I am using a BDM Interface. So If I then try and re-set the
watchdog and debug registers using GDB, GDB thinks they have already
been set and refuses to change them.

I Would imagine there are other scenarios people would get using
different embedded systems.

I was actually working on this as well, but I want to add an extra
function to allow the cache style of registers to be set. ie, "No
Cache", "Read Cached but never write cached", "Write Cached but never
Read Cached" and "Full Caching".

I was thinking about using a set command to do this, like:

set regcache $regname NOCACHE
set regcache $regname READCACHE
set regcache $regname FULLCACHE
set regcache $regname WRITECACHE
set regcacge all NOCACHE

The last one would allow you to set the cache style for all registers at
once.

I was also thinking that if this was a command then it would be
desirable to have:

info regcache 

which would print all the current cache settings.

maybe the cache flush would be better (in the fullness of time when all
things come to pass) as a:

set regcache $regname FLUSH
set regcache all FLUSH

to keep all this stuff contained together. 

What do you think?

I was just about to do some work in this area and I don't want to step
on anyones toes if they are also working in this direction. 

Also, do you know how one goes about turning $regname into a register
number? I haven't found this yet?

Steven Johnson.