Re: --diff-cmd arguments
From: Julian Foad <julianfoad_at_btopenworld.com>
Date: 2004-10-03 14:01:32 CEST
Chris Pepper wrote:
Yes, it's an ugly state of affairs, isn't it?
The work-around for the time being is to write a wrapper script that accepts the arguments as supplied (including '-u' and '-L label1' and '-L label2') and ignores some of them and calls the desired diff program. The arguments passed to the wrapper are, in this order:
+ one or more options (see below);
The initial options are the options specified in "--extensions" if that is non-blank, otherwise "-u".
(That may be a bug. I thought the rule ought to be: the options specified in "--extensions" if that is specified, otherwise "-u".)
Of course this ought to be documented somewhere, but is seems not to be.
> I was unsuccessful at finding the -u option in the svn source, but
I thought that used to work (used to prevent '-u' from being passed, though still passing two '-L label' options).
> Am I smoking crack? Is --extensions broken? It's certainly
Well, maybe a little bit broken. More like "we know it's not very good but haven't invented a better way of doing it". Note that "--extensions" is listed under "svn help diff" with a very short description. It also appears in the list of options for "svn diff" in the book, and is described in the "svn Switches" section of the book, but not described fully (the stuff about wrapper scripts and "-u" and "-L" isn't mentioned).
> a footnote in the book). I will try to address the doc aspect a bit,
Yes, a description of the current state of affairs is really needed. Your help with that would be very welcome.
- Julian
---------------------------------------------------------------------
|
This is an archived mail posted to the Subversion Dev mailing list.
This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.