--git revisited (was Re: mq: hg qnew (and the hgignore too?))

[Peter Arrenbrecht]

On Mon, Apr 28, 2008 at 5:57 PM, Marcin Kasperski
<Marcin.Kasperski at softax.com.pl> wrote:
> > Because the diff format has pretty much always been defined as
>  > whatever GNU patch accepts, and patch doesn't accept git diffs.
>
>  This is frequent excuse, but ... I feel the current default is wrong.
>  Compare:
>
>   [I used --git in spite I should not]
>
>   I send the patch to somebody, he rejects this patch. So I must
>   regenerate the diff with correct settings. Some time lost.
>
>   [I did not use --git in spite I should]
>
>   Binary files and permisions are dead lost.
>
>  What is more serious?
>
>  Note also that a lot of MQ usage scenarios does not require sending
>  patches at all (changeset editing for the simple example), or allow
>  for git patches...

And even when contributing patches, I use mq to develop but `hg email`
to send patches.

>  IMO it would be best if MQ used --git patches *always*, but provided
>  separate command to export the patch as old-style GNU patch (warning
>  about stripped info, if any).

What if mq issued warnings when encountering binary files/non-std
permissions in non-git mode?