Putting this in ~/.darcs/defaults should work but has no effect at all:

pull disable-ssh-cm
push disable-ssh-cm
put disable-ssh-cm
get disable-ssh-cm

Very annoying since Darcs's use of connection caching is pretty much  
broken on MacIntel, and the --disable-ssh-cm option is required to  
work around the problem.

Alexander.

On Wed, Nov 15, 2006 at 05:35:23PM +0000, Alexander Staubo wrote:
> Putting this in ~/.darcs/defaults should work but has no effect at all:
> 
> pull disable-ssh-cm
> push disable-ssh-cm
> put disable-ssh-cm
> get disable-ssh-cm

Can you try another option such as

whatsnew summary

to see whether this is specific to disable-ssh-cm or is a failure to parse
your defaults file?
-- 
David Roundy
Deptartment of Physics
Oregon State University

On Nov 15, 2006, at 21:22 , David Roundy wrote:

> David Roundy <droundy@darcs.net> added the comment:
>
> On Wed, Nov 15, 2006 at 05:35:23PM +0000, Alexander Staubo wrote:
>> Putting this in ~/.darcs/defaults should work but has no effect at  
>> all:
>>
>> pull disable-ssh-cm
>> push disable-ssh-cm
>> put disable-ssh-cm
>> get disable-ssh-cm
>
> Can you try another option such as
>
> whatsnew summary
>
> to see whether this is specific to disable-ssh-cm or is a failure  
> to parse
> your defaults file?

Yep, that works fine.

Alexander.

On Wed, Nov 15, 2006 at 17:35:23 +0000, Alexander Staubo wrote:
> pull disable-ssh-cm
> push disable-ssh-cm
> put disable-ssh-cm
> get disable-ssh-cm

disable-ssh-cm is actually broken in general in 1.0.8, but is fixed in
1.0.9 onwards

> Very annoying since Darcs's use of connection caching is pretty much  
> broken on MacIntel,

Interesting.  Can you tell us more?  I use it, and it seems fine,
although it doesn't do a very job cleaning up connections on a bad exit.
I'll have to think of a way to fix _that_; for now, a workaround might
consist of something like killall ssh and removing /tmp/darcs-ssh.

On Nov 15, 2006, at 23:01 , Eric Kow wrote:

> Eric Kow <eric.kow@gmail.com> added the comment:
>
> On Wed, Nov 15, 2006 at 17:35:23 +0000, Alexander Staubo wrote:
>> pull disable-ssh-cm
>> push disable-ssh-cm
>> put disable-ssh-cm
>> get disable-ssh-cm
>
> disable-ssh-cm is actually broken in general in 1.0.8, but is fixed in
> 1.0.9 onwards

I should have mentioned that this is on 1.0.9rc1, compiled using GHC  
6.6 for Intel.

>> Very annoying since Darcs's use of connection caching is pretty much
>> broken on MacIntel,
>
> Interesting.  Can you tell us more?  I use it, and it seems fine,
> although it doesn't do a very job cleaning up connections on a bad  
> exit.
> I'll have to think of a way to fix _that_; for now, a workaround might
> consist of something like killall ssh and removing /tmp/darcs-ssh.

The behaviour depends on the repo (small repos tend not to exhibit  
this problem), but it's almost consistently reproducible. From what I  
can tell, Darcs hangs waiting for a subprocess such as scp or sftp to  
complete. Sometimes Ctrl+C-ing Darcs will, inexplicably, kill the  
*subprocess* and let Darcs continue; sometimes Darcs will then  
continue successfully, but usually Darcs ends up with some partially  
read data that it will barf on.

I have seen the occasional orphaned SSH process left behind by Darcs,  
but when trying to beat Darcs into submission I have always made sure  
that no such orphans or /tmp garbage is around.

This is happening to everyone on my development team -- three  
different MacBooks, all on MacPorts + latest Tiger + SSHKeyChain --  
so it's not just me. Happens on pull, push, get and put. SSHKeyChain  
is not the most reliable of apps, though, and it may make sense to  
use ssh-agent on its own to eliminate one possible cause. I will also  
try to get some ktrace output.

Alexander.

fixed in 1.0.9