Need advice how to keep local changes without committingtomain branch

[Andrei Vermel]

----- Original Message ----- 
From: "Brendan Cully" <brendan at kublai.com>
To: "Andrei Vermel" <avermel at mail.ru>
Cc: <mercurial at selenic.com>
Sent: Thursday, December 21, 2006 10:57 PM
Subject: Re: Need advice how to keep local changes without committingtomain branch


> On Thursday, 21 December 2006 at 18:34, Andrei Vermel wrote:
>> 
>> From: "Brendan Cully" <brendan at kublai.com>
>> To: "Andrei Vermel" <avermel at mail.ru>
>> Cc: <mercurial at selenic.com>
>> Sent: Wednesday, December 13, 2006 7:36 PM
>> Subject: Re: Need advice how to keep local changes without committing 
>> tomain branch
>> 
>> Do you think it would be possible to somehow either optionally restore 
>> timestamp
>> of patched files, or allow commit/update of files that do not include the 
>> patched
>> files?
> 
> I'm not sure what you mean by the latter. You can qrefresh with -I/-X
> and/or a file list to bring only a subset of the working directory
> changes into a patch, leaving the rest as uncommitted
> modifications. But then you'd be doing a lot of qpush/qpop -f, which
> is probably dangerous.
> 
> You can also export a subset of your applied patches (eg hg export
> qbase:last-upstream). This is probably easiest when your local mods
> are at the top of the queue.
What I mean is a bit different. 
Suppose I have local changes in file1.txt and file2.txt, and these
are in an mq patch. 

Now I edit file3.txt and file4.txt, and I want to commit these upstream.
It would be handy to be able to just commit these as a regular hg changeset
without doing any pops/pushes. No idea if this is not possible or hard to 
implement, just a fantasy. I hope the fact that none of the files from the mq patch
(file1.txt, file2.txt) need to be committed may make this possible.
If file1.txt or file2.txt is asked to be committed, it should just fail.

Likewise suppose file3.txt, file4.txt change upstream, but none of
file1.txt and file2.txt. It would be handy to be able to just update without
pops or pushes.

> 
> Personally, I use ccache to make the rebuild after qselect --reapply
> -n fairly snappy.
Last time we hacked ccache to work with msvc, the combination
was too slow, probably because msvc did too much work before 
exporting source with all includes substituted. 
Does anyone know of ccache working well on windows?