"Eric Y. Kow" <[EMAIL PROTECTED]> writes:

>> > What about DARCS_SSH and the other commands using exec?
>> 
>> Particularly Eric might have an opinion, as he's borne the brunt of quite a
>> few ssh-related bug reports...
>
> Eek?
>
> Well, I'm mostly waiting on enlightenment on the DARCS_ARGUMENT thing
> first.  I also wonder if we would ever get into the situation where
> we try launch two commands at the same time and have them accidentally
> step over each others DARCS_ARGUMENTS toes.  I confess I really don't
> understand this stuff very well and can only say silly superficial
> things like "let's avoid introducing any more weird ssh-calling bugs
> if we can".

I don't see a situation where it can happen for edit_file/view_file
and i don't think we're using threads anymore, but I agree that changing
the ssh code is just not worth it.
I think there aren't many people using DARCS_SSH on unix and it's
much less important to handle every possible shell trick there.

One thing that we should support (i don't know if it works now) is
DARCS_SSH="C:\My Programs\putty\ssh -v", but i don't know how
windows handles spaces in filenames.

> If anything, I'd like to see the exec code using System.Process the
> basic principle being to keep our code simple and farm out all the ugly
> to dedicated libraries by people who've been thinking harder about this
> stuff than us (plus the Windows code uses it now, so it could be
> simpler).  But if we did this, would we lose Magnus Jonsson's rigourous
> error checking?

What about 6.2 support, can we use 6.4.1 (i remember some import bugs in
6.4) as oldest supported compiler for unstable?

Benedikt


_______________________________________________
darcs-devel mailing list
[email protected]
http://www.abridgegame.org/cgi-bin/mailman/listinfo/darcs-devel

Reply via email to