[PATCH 1 of 2] filemerge: add support for multiple merge tool versions (RFC)

Mathias De Maré mathias.demare at gmail.com
Wed Jan 6 06:57:16 UTC 2016


On Tue, Dec 22, 2015 at 8:06 PM, Matt Mackall <mpm at selenic.com> wrote:

> On Tue, 2015-12-22 at 19:34 +0100, Mathias De Maré wrote:
> > On Mon, Dec 21, 2015 at 2:29 PM, Yuya Nishihara <yuya at tcha.org> wrote:
> >
> > > On Sun, 20 Dec 2015 08:48:49 +0100, Mathias De Maré wrote:
> > > > On Fri, Dec 18, 2015 at 8:28 PM, Matt Mackall <mpm at selenic.com>
> wrote:
> > > > > On Wed, 2015-12-16 at 01:07 -0600, Mathias De Maré wrote:
> > > > > > # HG changeset patch
> > > > > > # User Mathias De Maré <mathias.demare at gmail.com>
> > > > > > # Date 1450249051 -3600
> > > > > > #      Wed Dec 16 07:57:31 2015 +0100
> > > > > > # Node ID 958659a81ec2cea8037e4e8ba1bf9841c0272f88
> > > > > > # Parent  7e8a883da1711008262150bb6f64131812de3e0b
> > > > > > filemerge: add support for multiple merge tool versions (RFC)
> > > > >
> > > > > ENODOCS. But from trying to parse the test, it seems a little
> complex.
> > > > > It'd be much simpler to have something like:
> > > > >
> > > > > mytoolwithauto = /usr/bin/mytool
> > > > > mytoolwithauto.featurecheck = "--help", ".*--auto-merge.*"
> > > > >
> > > > Hm, that does seem a lot less complex. I'll have a look if I can
> create
> > > > something like this (and document it). Thanks!
> > >
> > > FWIW, it won't be useful on Windows because GUI applications can't
> access
> > > to
> > > stdout in general. For example, "kdiff3 --confighelp" opens GUI on
> Windows.
> > >
> > 'meld --help' doesn't print anything and doesn't launch a GUI either.
> >
> > Would it be acceptable to simply disable the feature check on Windows?
> That
> > would mean the experience on Windows stays the same, but the experience
> on
> > other OS'es improves.
>
> I'd rather not. Similar issues probably exist for some Mac and Linux
> editors
> because kids today don't do Unix good. And it forecloses feature detection
> with
> more sensible tools.
>
> Another strategy would be to try running with the option and looking at
> output/exit codes.
>

I looked into this and it seems it would add way too much complexity.
The main reason not to add '--auto-merge' (if I read it correctly in the
old threads) is concern about custom packages for CentOS 5 and 6 breaking.
I could extend buildrpm (or dockerrpm) to create a file
/etc/mercurial/hgrc.d/mergetools.centos[5,6].rc containing only the old
arguments for Meld. All other OSes would get the regular arguments (which
we can change to add '--auto-merge'). I believe this would be an easier
approach that does not introduce too much complexity.

Does this seem like a good approach?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20160106/3b3422d5/attachment.html>


More information about the Mercurial-devel mailing list