Re: win2k/cygwin cannot handle even moderately sized packs

From: Shawn Pearce <spearce@spearce.org>
Date: Nov 7, 2006, 18:56 UTC
Message-ID: <20061107185648.GE26591@spearce.org>
In-Reply-To: <20061107182636.GD26591@spearce.org>

Shawn Pearce <spearce@spearce.org> wrote:

Show 16 quoted lines

> Shawn Pearce <spearce@spearce.org> wrote:
> > Alex Riesen <raa.lkml@gmail.com> wrote:
> > > >The patches are on the mailing list archives somewhere around
> > > >Sept. 5th timeframe from me; as I recall we dropped them as they
> > > >didn't apply on top of Junio's 64 bit index changes (which were
> > > >reverted out of next anyway).
> > > 
> > > I seem to be unable to find them. Does anyone still has the
> > > patches/branch please? Junio, you did sound interested?
> > > (God, I wish I have paid attention then...)
> 
> You can't find them because I never sent them.  *sigh*
> 
> Too f'ing bad this .patch I created with format-patch doesn't say
> what commits it was based on, 'cause I can't find anything it will
> apply too.  Or better, too f'ing bad I deleted that branch.  *sigh*

This thread has now proven without any shadow of a doubt that I can be an idiot sometimes.

In this case my patch didn't apply because it was 3rd in a series; I had two of those patches but lacked the third (a simple cleanup patch). Redoing that simple cleanup let everything apply.

I've pushed the changes out to repo.or.cz:

	http://repo.or.cz/w/git/fastimport.git

in the window-mapping branch. Note that this is based on a slightly older version of Git (v1.4.2). There are two "tuneables" on line 376 of sha1_file.c, this is the maximum amount of memory (in bytes) to denote to packs and the maximum chunk size of each pack (in bytes). I planned on making these configuration options but didn't get to that yet.