Re: [PATCH 1/1] Use correct /dev/null for UNIX and Windows

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi Junio,

On Wed, 31 Oct 2018, Junio C Hamano wrote:

> Chris Webster <chris@webstech.net> writes:
> 
> >>> > Use File::Spec->devnull() for output redirection to avoid messages
> >>> > when Windows version of Perl is first in path.  The message 'The
> >>>
> >>> Dscho, "Windows version of Perl is first in path" somehow feels
> >>> contradicting with what one of the topics I saw from you were trying
> >>> to enforce (or, at least, "set as the supported configuration").
> >>>
> >>> I am guessing that the Perl you are building and shipping with Git
> >>> for Windows would yield what the shell that ends up running the
> >>> scriptlet `git config --get-color $key` prefers when asked for
> >>> File::Spec->devnull(), and nothing will break with this patch even
> >>> if that is "/dev/null", but I thought I'd double check.
> >>>
> >>> Thanks.
> >>>
> > This problem originally showed up in the
> > https://github.com/so-fancy/diff-so-fancy project, which has a copy of
> > DiffHighlight.pm.   That project allows diffsofancy (perl) to be run
> > from the command line without requiring the bash environment ((well ,
> > sort of) including the associated perl).
> 
> Thanks for additional comments.  
> 
> In any case, Windows is not my bailiwick, so I'll hope that the
> above comments from you would help Dscho in his response and wait.
> I know use of File::Spec->devnull() won't hurt POSIX folks so making
> sure this won't break Git for Windows is the primary thing I woudl
> worry about this patch.

Indeed, the patch in question regards something I consider outside Git for
Windows' realm. As Chris said, you can run this script from a PowerShell
prompt, without any Git Bash (and without Git's Perl) involved.

I am fine with this patch, as long as the author name is fixed to match
the name in the Signed-off-by: footer ;-) [*1*]

Ciao,
Dscho

Footnote *1*: This patch came in via GitGitGadget, and if I had infinite
amounts of time, I would probably implement some rudimentary pre-checks,
such as: does the Author: header match the first Signed-off-by: footer, is
the commit message wrapped correctly, does the oneline have a prefix and
continues lower-case, etc. And GitGitGadget would then point out the
issues, possibly even try to fix them and push up the fixed commits.

If anybody agrees with these goals and is curious enough to dive into some
Typescript programming, I'd be very happy to guide that person through
implementing this ;-)