Stephen Leake, 2007-04-28:
> Christian Ohler <[EMAIL PROTECTED]> writes:
>
>> A simple test to check the basic named pipe features that xmtn needs
>> (not full standards compliance) would be this: Open a terminal window
>> and type
>>
>> mkfifo a
>> cat a
>>
>> Open another terminal window and type
>>
>> cat >a
>>
>> and then type a few lines of text. Each line should appear in the
>> first terminal window as you enter it. When you hit ctrl+d on an
>> empty line, both cat processes should terminate.
>
> Ok. That works with Cygwin.
>
> MinGW doesn't have mkfifo.
Thanks for the information.
> I'm happy using Cygwin, but a lot of people are not. And since
> both monotone and DVC can otherwise live with MinGW, I don't think we
> can require Cygwin just for this.
Agreed.
An even simpler hack around this would be redirecting stderr to a
regular file (as we do now) and running tail -f -c +0 on it. Does this
work in Cygwin and MinGW? Maybe this is what we should be doing.
> I have not dug into the xmtn code enough to understand why it uses
> automate stdio sessions instead of single automate commands. Can you
> explain?
See the comments at the top of xmtn-automate.el. Basically, it's a
potential performance advantage when sending a large number of commands
to monotone. In addition to avoiding the overhead of spawning a new
process for each command, a persistent mtn process can potentially do
some interesting caching. ("Persistent" for one sequence of commands,
not for an indefinite period.)
xmtn can send large numbers of commands to monotone while building
revlist buffers and while resolving revision IDs, and maybe during other
operations.
automate stdio also allows us to pass arbitrary binary data to monotone
in a well-defined way; this is less easy with command-line arguments.
(Filenames with non-ASCII characters come to mind.)
Christian.
_______________________________________________
Dvc-dev mailing list
[email protected]
https://mail.gna.org/listinfo/dvc-dev
