[ros-dev] Cache / Memory Manager / FileSystemDrivers

[art yerkes]

On Wed, 7 Jan 2009 20:04:34 +0300
Aleksey Bragin <aleksey at reactos.org> wrote:

> On Jan 7, 2009, at 7:24 PM, WaxDragon wrote:
> 
> > On Wed, Jan 7, 2009 at 9:18 AM, Aleksey Bragin  
> > <aleksey at reactos.org> wrote:
> >>        Hello,
> >> I recently spent some time experimenting with various Cc rewrites we
> >> have in our tree, and before that I extensively tested our fastfat
> >> driver in a real NT environment (MS Windows 2003).
> >>
> >> The conclusion of the above work is that our fastfat driver either
> >> needs serious bugfixing, or a rewrite is needed. It corrupts
> >> directory tables in a real NT, leads to unusual behaviour of cc-
> >> rewrite branch, and may have side effects on arty's newcc.
> >>
> >> Before this is done (bugfixing or rewriting, developing and testing
> >> against a Windows 2003 at least, not ReactOS), it's meaningless to
> >> continue any other work on Cc and related Mm parts.
> >>
> >>
> >> WBR,
> >> Aleksey Bragin.
> >
> > Let's make a plan!  I want a proper branch for the cc rewrites.  I'd
> > like to have a live branch, not this half branch, half patch form that
> > cc_rewrite is in now.  Perhaps this would be a good place for the
> > working fastfat to be dropped when finished.
> >
> > I have time and resources that I could put towards this goal.  Let's
> > get fastfat working on 2k3, then drop it into a new cc branch and get
> > it working..
> >
> > <Leroy Jenkins>LET'S DO THIS, CHUMS!</Leroy Jenkins>
> >
> > WD
> 
> Live branch needs constant merging. Have a look at amd64 bringup  
> branch: most of commits are merges. That mm.patch is an ugly way of  
> keeping the branch always uptodate...
> 
> Anyway, I have made some fixes to the branch, so if someone wants to  
> test, I would need to commit my changes.
> 
> As for the fastfat, I think the problems is in synchronization  
> somewhere, not in a pool corruption, passing invalid parameters,  
> misusing Cc API, or anything like that - I fixed that already.
> 
> Another idea was to take ext2fsd by Matt Wu, which is supposed to  
> work, and convert it to a fat32/16/12 driver.

I've been working on some things under git over svn.  Merging trunk
has been pretty easy this way, although some assembly was required.
Perhaps rather than switching, we can start a git repository just
for branches and feed it from svn, also using it to build diffs for
merging back to trunk.
-- 
art yerkes <ayerkes at speakeasy.net>