Re: [PATCH 5/8] Teach run_command how to setup a stdin pipe

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Mar 11, 2007, 02:07 UTC
Message-ID: <Pine.LNX.4.63.0703110305000.22628@wbgn013.biozentrum.uni-wuerzburg.de>
In-Reply-To: <20070311014533.GA10343@spearce.org>

Hi,

On Sat, 10 Mar 2007, Shawn O. Pearce wrote:

Show 21 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> wrote:
> > On Sat, 10 Mar 2007, Shawn O. Pearce wrote:
> > 
> > > Sometimes callers trying to use run_command to execute a child 
> > > process will want to setup a pipe or file descriptor to redirect 
> > > into the child's stdin.
> > > 
> > > This idea is completely stolen from builtin-bundle's fork_with_pipe, 
> > > written by Johannes Schindelin.  All credit (and blame) should lie 
> > > with Dscho.  ;-)
> > 
> > ;-)
> > 
> > Thank you for starting this thread. I think it makes a lot of sense in 
> > the face of the MinGW port. (I am not interested in the hook stuff 
> > personally, so I'll not comment on that.)
> 
> I'd like to see run_command learn how to also redirect stdout, then 
> replace fork_with_pipe in builtin-bundle with run_command. We should be 
> able to also improve some of our other more direct uses of fork to use 
> run_command at that point too.

I concur.

> I don't know squat about MinGW, so I don't know if having the 
> run_command abstraction really helps there or not, but I know we tried 
> to make it with the good intentions of being able to use it to wrap that 
> asinine CreateProcess() thing that Win32 has...

There is only one place that I am aware of, where this is not enough: For the shallow thing, upload-pack fork()s but does not execute another program, but lists all the objects to be uploaded. (For shallow, this _cannot_ be an external process, since it mucks with the commits which are to be treated as shallow).

Ciao, Dscho