when darcs fails to create a SSH connections (eg. due to firewall limits on
connections) it will complain

darcs: failed to read patch in get_extra:

Perhaps this is a 'partial' repository? 


This diagnostic is misleading; can have a more detailed reason why 'get_patch
failed' ? 

thanks!

On Sun, Nov 27, 2005 at 10:39:30AM +0000, Han-Wen Nienhuys wrote:
> when darcs fails to create a SSH connections (eg. due to firewall limits on
> connections) it will complain
> 
> darcs: failed to read patch in get_extra:
> 
> Perhaps this is a 'partial' repository? 
> 
> 
> This diagnostic is misleading; can have a more detailed reason why
> 'get_patch failed' ?

Hmmm.  The reason for the lack of an informative error message is that
we're using a Maybe Patch (which is either a patch or Nothing) to indicate
whether the read patch succeeded.  We could switch to an Either String
Patch, which will allow us to give a useful error message (provided ssh
gives us a useful error message).  This would be a pretty pervasive change,
but should be pretty safe--as the type system will force us to make it
everywhere.
-- 
David Roundy
http://www.darcs.net

I'd be happy to look into this.

I redefined "get_extra_old" which appears to be the only place that error
message can be generated from.  I turned it into "undefined", ran the test suite
and everything passed.  This tells me we don't have any test cases that test
this problem.  I guess my first step will be to figure out a reliable test case.

On Sat, Dec 31, 2005 at 02:02:59AM +0000, Jason Dagit wrote:
> I redefined "get_extra_old" which appears to be the only place that error
> message can be generated from.  I turned it into "undefined", ran the
> test suite and everything passed.  This tells me we don't have any test
> cases that test this problem.  I guess my first step will be to figure
> out a reliable test case.

I believe you could trigger this error message simply by deleting one of
the patch files in a local repository (created in the test suite).  This
would simulate some sort of connection failure that keeps us from accessing
that patch file.  You could also try making the patch file non-readable
(which might be more similar to a scenario that users would actually see).
-- 
David Roundy
http://www.darcs.net

On Dec 31, 2005, at 8:35 AM, David Roundy wrote:

>
> David Roundy <droundy@darcs.net> added the comment:
>
> On Sat, Dec 31, 2005 at 02:02:59AM +0000, Jason Dagit wrote:
>> I redefined "get_extra_old" which appears to be the only place  
>> that error
>> message can be generated from.  I turned it into "undefined", ran the
>> test suite and everything passed.  This tells me we don't have any  
>> test
>> cases that test this problem.  I guess my first step will be to  
>> figure
>> out a reliable test case.
>
> I believe you could trigger this error message simply by deleting  
> one of
> the patch files in a local repository (created in the test suite).   
> This
> would simulate some sort of connection failure that keeps us from  
> accessing
> that patch file.  You could also try making the patch file non- 
> readable
> (which might be more similar to a scenario that users would  
> actually see).

I tried both of those actually and neither did the trick  :)  I also  
discovered that this could be caused in patchset_unison or possibly  
when dealing with partial repos.  I'm still trying to decipher some  
code where it could be happening, but I'm not sure if I'll be able to  
fix this.  Either way, the test suite is inadequate in this area.  I  
should try that test again since I have 'undefined' several more  
functions which call get_extra_old.

Also worth noting, copySSH does not print the scp error message, but  
copySSHs does.  Perhaps I should just modify copySSH to be more like  
copySSHs in that regard.  Even if I do that I would like to improve  
the test suite so I can see the output and verify my changes.

Thanks,
Jason

I believed darcs "get" provides much better feedback on failure in the unstable
branch now. 

Please re-open if you find that the problem persists.