Async patch (no. 3)

Async patch (no. 3)

[Nick Roberts]

[-- Attachment #1: message body and .signature --]
[-- Type: text/plain, Size: 1075 bytes --]


This async patch gives the same test results for synchronous and asynchronous
operation.  Ignoring sigstep.exp again, I get:

		=== gdb Summary ===

# of expected passes		10851
# of unexpected failures	44
# of unexpected successes	1
# of expected failures		42
# of unknown successes		3
# of known failures		66
# of untested testcases		5
# of unsupported tests		5

for both modes.  I also get the same result for current mainline GDB in CVS
(although in this case I sometimes get extra fails for schedlock.exp).

Currently it only works for i686-pc-linux-gnu and I've tried to hide it from
other architectures/OSes.  The only visible change should be those made to MI.

If the global maintainers are agreeable, in principle, to it's inclusion after
the next release, I will tidy it up and provide a ChangeLog.  I would
appreciate any clues as to where some things should go e.g waitpid in
event-loop.c to linux-nat.c? (How?)

This patch will make the branch nickrob-async-20060513 defunct.

-- 
Nick                                           http://www.inet.net.nz/~nickrob


[-- Attachment #2: Asynchronous GDB --]
[-- Type: application/octet-stream, Size: 43891 bytes --]

Re: Async patch (no. 3)

[Daniel Jacobowitz]

On Fri, Oct 13, 2006 at 09:33:14PM +1300, Nick Roberts wrote:
> If the global maintainers are agreeable, in principle, to it's inclusion after
> the next release, I will tidy it up and provide a ChangeLog.  I would
> appreciate any clues as to where some things should go e.g waitpid in
> event-loop.c to linux-nat.c? (How?)

I'm agreeable in principle, though I don't have time to review it as
well as I'd like at the moment.  Thank you for all your hard work on
this!  I'll try hard to make the time.

I think you will need to add new target_ops methods; probably one to
set up async waiting and another to clean up afterwards.  There
shouldn't be any reference to SIGCHLD directly in the common files (I'm
thinking about how to make this work on Windows, you see).

-- 
Daniel Jacobowitz
CodeSourcery

Re: Async patch (no. 3)

[Nick Roberts]

 > I'm agreeable in principle, though I don't have time to review it as
 > well as I'd like at the moment.  Thank you for all your hard work on
 > this!  I'll try hard to make the time.

Thanks.  There's no great urgency on my part, I just need it for after Emacs
release.  It's about to go to pretest but this will probably be a drawn out
process if the two and a half year feature freeze is anything to go by.

 > I think you will need to add new target_ops methods; probably one to
 > set up async waiting and another to clean up afterwards.  There
 > shouldn't be any reference to SIGCHLD directly in the common files (I'm
 > thinking about how to make this work on Windows, you see).

OK, I'll try that.

In the meantime, I've noticed that the patch forces GDB to run synchronously.
In target.c, update_current_target must be made to inherit async methods again.
This means that server-run.exp still doesn't work (and probably any test which
changes to a synchronous target).  I'll include this when I submit a cleaned up
patch.


-- 
Nick                                           http://www.inet.net.nz/~nickrob