Re: [EGIT PATCH 00/20] PackWriter, first usable attempt

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Marek Zawirski <marek.zawirski@xxxxxxxxx> wrote:
> At first, some stuff was still missing to produce packs, mostly
> raw-data access related and ObjectWalk related.

I'm glad it turned out to be so little missing actually.  Reusing
ObjectWalk saved a lot of code in the pack writer, and for the most
part our existing data access structures were already well organized.

It is too early to say how the performance is going to work, but
object packing with delta reuse can be difficult and I'm happy to
see that our abstractions more-or-less supported it.  Tuning can
come later, once we better understand the code, and have something
for end-users to complain (or praise) about.
 
> Finally, we've got some support for pack writing! It's not that
> power that C git version offers, but something usable. Delta
> generation is not supported. Although we can reuse deltas and objects,
> and support all other (I hope) options of git-pack-objects directly or
> indirectly, most importantly --thin.
> 
> Pack writing and some other features are tested, seem to work.
> 
> This implementation of packing is not a very valuable thing directly
> (achieving efficient storage), however it's a base for enhancements
> and can be used for sending packs over net (with some assumptions).
> It's more a "repacking" than "packing" tool.

Yup.  The critical part here is jgit can now format a pack file,
which means we can now actually implement native push over the
local pipe (to fork+exec'd git-receive-pack) or over SSH.  That
is one of the major missing features in the Eclipse plugin, so
this is a huge milestone for us.  Thank you Marek.

> So... I'm switching now to push implementation. If time allows,
> delta-algorithms will be added later.

Yay.

Native push protocol support at this point is much more important
than delta generation.  Although delta generation is one of the
key features that makes git so damn efficient it is pointless if we
cannot actually communicate with a remote repository to send them
our changes.  Early adopters of the push support coming from this
plugin can at least use it on local area networks, where bandwidth
is not (usually) a limiting factor.

>  28 files changed, 2258 insertions(+), 73 deletions(-)

Nice to see it didn't take that much code either.

-- 
Shawn.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux