exact meaning of restart

exact meaning of restart

[Quality Quorum]

Hi, 

Say I have embedded system, it got 'R' request, what it should do ?

I have RTEMS/pc386 bootstrapped by the netboot off the linux box.
Once it get 'R' it will do warm-boot. It seems like it gdb 
expects something else:

1. Gdb does not allow enough time to complete reboot.
2. Gdb does not repeat '!' request, so system could 
   be rebooted remotely only once.

I am not able to figure out what is required in the case.

My only guess is that extended operations are limited to 
gdbserver like environments which allow fast restart and
embedded systems should use target 'remote' and reboot themselves 
once 'kill' is received.

Can anybody help ?

Thanks,

Aleksey

Re: exact meaning of restart

[Quality Quorum]

y

On Sat, 18 Dec 1999, Andrew Cagney wrote:

> Quality Quorum wrote:
> > 
> > Hi,
> > 
> > Say I have embedded system, it got 'R' request, what it should do ?
> > 
> > I have RTEMS/pc386 bootstrapped by the netboot off the linux box.
> > Once it get 'R' it will do warm-boot. It seems like it gdb
> > expects something else:
> > 
> > 1. Gdb does not allow enough time to complete reboot.
> > 2. Gdb does not repeat '!' request, so system could
> >    be rebooted remotely only once.
> > 
> > I am not able to figure out what is required in the case.
> > 
> > My only guess is that extended operations are limited to
> > gdbserver like environments which allow fast restart and
> > embedded systems should use target 'remote' and reboot themselves
> > once 'kill' is received.
> > 
> > Can anybody help ?
> 
> Yes.  As far as is known ``!'' is limited to the GDB server.

And gdbserver (1)returns status string in response to '!',
(2)is not programmed in ROM  - so we do not need 'forcextended'
to provide backward compatibility with old stubs - life is not
so bad and it seesm really trivial feat to fix remote.c to 
seamlessly support extended ops on both gdbserver (with its
fast reload) and embedded targets (wit its slow reload).

Thanks a lot.
> 
> 	Andrew
> 


Thanks,

Aleksey

Re: exact meaning of restart

[Andrew Cagney]

Quality Quorum wrote:
> 
> y
> 
> On Sat, 18 Dec 1999, Andrew Cagney wrote:
> 
> > Quality Quorum wrote:
> > >
> > > Hi,
> > >
> > > Say I have embedded system, it got 'R' request, what it should do ?
> > >
> > > I have RTEMS/pc386 bootstrapped by the netboot off the linux box.
> > > Once it get 'R' it will do warm-boot. It seems like it gdb
> > > expects something else:
> > >
> > > 1. Gdb does not allow enough time to complete reboot.
> > > 2. Gdb does not repeat '!' request, so system could
> > >    be rebooted remotely only once.
> > >
> > > I am not able to figure out what is required in the case.
> > >
> > > My only guess is that extended operations are limited to
> > > gdbserver like environments which allow fast restart and
> > > embedded systems should use target 'remote' and reboot themselves
> > > once 'kill' is received.
> > >
> > > Can anybody help ?
> >
> > Yes.  As far as is known ``!'' is limited to the GDB server.
> 
> And gdbserver (1)returns status string in response to '!',
> (2)is not programmed in ROM  - so we do not need 'forcextended'
> to provide backward compatibility with old stubs - life is not
> so bad and it seesm really trivial feat to fix remote.c to
> seamlessly support extended ops on both gdbserver (with its
> fast reload) and embedded targets (wit its slow reload).
> 
> Thanks a lot.
> >
> >       Andrew
> >

FYI,

My opinion on this is that the extended operator ``!'' should go away. 
Having one command alter the semantics of several others isn't right. 
The relevant commands, and remote.c's use of them should be tightend to
the point that the operator is no longer needed.

	Andrew
From qqi@world.std.com Thu Dec 23 17:38:00 1999
From: Quality Quorum <qqi@world.std.com>
To: Andrew Cagney <ac131313@cygnus.com>
Cc: gdb@sourceware.cygnus.com
Subject: Re: exact meaning of restart
Date: Thu, 23 Dec 1999 17:38:00 -0000
Message-id: <Pine.SGI.3.95.991223203519.13428C-100000@world.std.com>
References: <38629C06.19C3B0F1@cygnus.com>
X-SW-Source: 1999-q4/msg00553.html
Content-length: 676

On Fri, 24 Dec 1999, Andrew Cagney wrote:

> 
> My opinion on this is that the extended operator ``!'' should go away. 
> Having one command alter the semantics of several others isn't right. 

IMHO, it is inevitable: we have gorups of commands which do not have
sense separate from each other.

> The relevant commands, and remote.c's use of them should be tightend to
> the point that the operator is no longer needed.

'!' does just two things allows 'restart' and changes semantics of kill
in some cases - it does not seem like a big problem too me and
funtionality is valuable in my view. I would not have as separate
tharget though.

> 
> 	Andrew
> 

Thanks,

Aleksey