Re: [PATCH 0/3] mingw: introduce a way to avoid std handle inheritance

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Oct 31, 2017, 17:12 UTC
Message-ID: <alpine.DEB.2.21.1.1710311809570.6482@virtualbox>
In-Reply-To: <20171030205522.hur26cumwusk7wwa@aiede.mtv.corp.google.com>

Hi Jonathan,

On Mon, 30 Oct 2017, Jonathan Nieder wrote:

Show 20 quoted lines

> Johannes Schindelin wrote:
> 
> > Particularly when calling Git from applications, such as Visual Studio,
> > it is important that stdin/stdout/stderr are closed properly. However,
> > when spawning processes on Windows, those handles must be marked as
> > inheritable if we want to use them, but that flag is a global flag and
> > may very well be used by other spawned processes which then do not know
> > to close those handles.
> >
> > As a workaround, introduce handling for the environment variables
> > GIT_REDIRECT_STD* to read/write from/to named pipes instead
> > (conceptually similar to Unix sockets, for you Linux folks). These do
> > not need to be marked as inheritable, as the process can simply open the
> > named pipe. No global flags. No problems.
> >
> > This feature was introduced as an experimental feature into Git for
> > Windows v2.11.0(2) and has been tested ever since. I feel it is
> > well-tested enough that it can be integrated into core Git.
> 
> Can this rationale go in the commit messages?

I thought I had done exactly that in 1/3...

> Actually I wouldn't mind if this were all a single patch, with such a
> rationale in the commit message.

Quite honestly, I'd rather not. There is a logical structure to the three patches (and I left a Duck in 3/3, to see who bothers to actually read what I wrote).

The redirection, for example, is a very special thing that I would like to keep in a separate commit, for clarity.

> The patches' concept seems sane.  I haven't looked closely at the
> implementation.

Thanks, Dscho