Skip to contentSkip to conversation

Linux v2.6.16-c2

DISCUSSION Larry Finger 5 participants 12 messages Feb 3, 2006 – Feb 4, 2006

lore
Share & tools
lore
#0 lore raw

I know you must have a good reason to switch to the 'pack' form in the git tree, but I'm curious. I did a git pull late yesterday, which was "normal", and another this morning when I saw that 2.6.16-rc2 was posted. I was quite surprised to download 110 MB of data to get roughly 150 changed lines.

Larry

#1 lore raw
Quoted 5 lines · Larry Finger

I know you must have a good reason to switch to the 'pack' form in the git tree, but I'm curious. I did a git pull late yesterday, which was "normal", and another this morning when I saw that 2.6.16-rc2 was posted. I was quite surprised to download 110 MB of data to get roughly 150 changed lines.

Don't use rsync (or http) access unless you have to.

Try using "git://git.kernel.org/" instead.

Now, it may be that if we have lots of people using it, the CPU usage of the server-side effort will go through the roof, and we'll have to come up with something else (most likely meaning having some mirrors also run git-daemon, so that the CPU overhead can be pushed out too).

The point being that you shouldn't see the packing as even an issue (it should be a per-repository decision). The fact that it shows up is because of using non-git-aware protocols.

Linus

#2 lore raw
Quoted 10 lines · Linus Torvalds
Quoted 4 lines · Larry Finger

I know you must have a good reason to switch to the 'pack' form in the git tree, but I'm curious. I did a git pull late yesterday, which was "normal", and another this morning when I saw that 2.6.16-rc2 was posted. I was quite surprised to download 110 MB of data to get roughly 150 changed lines.

Don't use rsync (or http) access unless you have to.

Quoted 8 lines

Try using "git://git.kernel.org/" instead.

Now, it may be that if we have lots of people using it, the CPU usage of the server-side effort will go through the roof, and we'll have to come up with something else (most likely meaning having some mirrors also run git-daemon, so that the CPU overhead can be pushed out too).

The point being that you shouldn't see the packing as even an issue (it should be a per-repository decision). The fact that it shows up is because of using non-git-aware protocols.

Linus

Thanks for the explanation. I have to admit that git is pretty much a black box to me. I use the guide at http://linux.yyz.us/git-howto.html and it recommends using rsync. I'll have to figure out how to change to git protocol.

Larry

#3 lore raw
Quoted 3 lines · On Fri, 03 Feb 2006 12:25:05 -0600 Larry Finger <[email protected]>…

Thanks for the explanation. I have to admit that git is pretty much a black box to me. I use the guide at http://linux.yyz.us/git-howto.html and it recommends using rsync. I'll have to figure out how to change to git protocol.

Just do:

git pull git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

:)

Signature · 2 lines
	Paolo Ornati
	Linux 2.6.15-kolivasPatch on x86_64
#4 lore raw
Quoted 4 lines · Paolo Ornati

On Fri, 03 Feb 2006 12:25:05 -0600
Larry Finger <[email protected]> wrote:

Quoted 9 lines
Quoted 1 line

Thanks for the explanation. I have to admit that git is pretty much a black box to me. I use the guide at http://linux.yyz.us/git-howto.html and it recommends using rsync. I'll have to figure out how to change to git protocol.

Just do:

git pull git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

:)

Got it now. Thanks.

Larry

#5 lore raw
Quoted 13 lines · Larry Finger
Quoted 12 lines · Paolo Ornati
Quoted 4 lines · On Fri, 03 Feb 2006 12:25:05 -0600 Larry Finger <[email protected]>…

Thanks for the explanation. I have to admit that git is pretty much a black box to me. I use the guide at http://linux.yyz.us/git-howto.html and it recommends using rsync. I'll have to figure out how to change to git protocol.

Just do:

Quoted 7 lines
Quoted 5 lines

git pull
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

:)

Got it now. Thanks.

You can also edit the file that describe your shorthand notation. If you normally do "git pull origin" just look into the ".git/remotes/origin" file, and I think you'll find it very obvious what it all does.

(If you used a really old version of git to create the archive originally, it might be ".git/branches/origin" instead).

Linus

#6 lore raw
Quoted 7 lines · Linus Torvalds

You can also edit the file that describe your shorthand notation. If you normally do "git pull origin" just look into the ".git/remotes/origin" file, and I think you'll find it very obvious what it all does.

(If you used a really old version of git to create the archive originally, it might be ".git/branches/origin" instead).

Linus

Thanks. I was almost ready to create aliases so that I would not have to remember all those paths. My initial copy of git was obviously new enough that .git/remotes/origin had the info. Now rid of rsync.

Thanks again,

Larry

#7 lore raw
Quoted 4 lines · Larry Finger

Thanks. I was almost ready to create aliases so that I would not have to remember all those paths. My initial copy of git was obviously new enough that .git/remotes/origin had the info. Now rid of rsync.

You may want to do a "git repack -a -d" to get everything condensed into a single pack file. It will likely take a while to run, however.

Signature · 2 lines
Ryan Anderson
  sometimes Pug Majere
#8 lore raw
Quoted 3 lines · Ryan Anderson

You may want to do a "git repack -a -d" to get everything condensed into a single pack file. It will likely take a while to run, however.

Yes. Also remember to throw out the unnecessary files afterwards with "git prune-packed". Otherwise you'll have tons of useless small files that contain all the same info that you already have in the packfile.

Linus

#9 lore raw
In reply to Ryan Anderson
Quoted 2 lines · Ryan Anderson

You may want to do a "git repack -a -d" to get everything condensed into a single pack file. It will likely take a while to run, however.

diego@estel ~/kernel/linux-2.6 # git repack -a -d Packing 185104 objects
Pack pack-b87f4fd87979fe91a0141c7037b3dfbddd0a8c0a created.
error: wrong index file size
diego@estel ~/kernel/linux-2.6 #

Is this expected? :/ (git 1.1.5)

#10 lore raw
Quoted 8 lines · Diego Calleja

diego@estel ~/kernel/linux-2.6 # git repack -a -d Packing 185104 objects
Pack pack-b87f4fd87979fe91a0141c7037b3dfbddd0a8c0a created.
error: wrong index file size
diego@estel ~/kernel/linux-2.6 #

Is this expected? :/ (git 1.1.5)

Absolutely not.

You have a corrupted pack somewhere. I suspect that you may have interrupted a "rsync" transfer, and have a partial pack-file (or, in this case, the index file _descibing_ that pack-file) in your repository as a result.

Did you perhaps do a ^C in frustration when you first did the rsync: pull, and noticed that it was going to pull the whole big new pack-file?

It's kind of as if you had copied a CVS archive by doing a "cp -r" on CVSROOT, and interrupted it half-way.

You can run

	git fsck-objects --full

which will do a full fsck and warn about anything it finds.

(and any pack it complains about you should just remove: if it turns out that the pack actually contained some data you need, at worst you can then use "rsync" to repopulate the .git object database, and while some object damage can be repaired, it's simply not _worth_ repairing it).

Linus

#11 lore raw
Quoted 2 lines · Linus Torvalds

Did you perhaps do a ^C in frustration when you first did the rsync: pull, and noticed that it was going to pull the whole big new pack-file?

I used the native git:// protocol. I'm on a dialup connection. Maybe the link disconnected while I was doing "git pull" and I didn't noticed it.

Quoted 3 lines
	git fsck-objects --full

which will do a full fsck and warn about anything it finds.

It seems that this repository is definitively corrupted, so I think I'll just download a new copy :/

broken link from    tree 26dcdcda6e1cb010730d3bd9caa3e5e9209be767
              to    blob f769872debea6e09f1511f1e86558ed4ce50af9e
broken link from    tree 26dcdcda6e1cb010730d3bd9caa3e5e9209be767
              to    blob a85e16f56d73672dce8de3c61a4ee7de68028c15
broken link from    tree 26dcdcda6e1cb010730d3bd9caa3e5e9209be767
              to    blob c631c082f8f7d281464f7eb8c5df07791e2836ab
broken link from    tree 26dcdcda6e1cb010730d3bd9caa3e5e9209be767
              to    blob f942fcc218312ec70d6eb9d1793ba43cd98ad355

missing blob 26dfa9e216c2814f9419b318ed8289e46f6b8a21 missing blob 26e421498c973dc3c539263d8d6d19431ef14bc2 missing blob 26eac194064b8a2851705db6cfa3bd1bd5bc5e8a missing tree 26eb17786227897888ff6447389e979f2e2f3586 missing blob 26ee81bbd6c64e4d2ace3ad5eec300b783facf0b missing blob 26f17e3fc45c6113b99e138dde03138502e85b4d missing blob 26f293ab96178e22034626adf5fdae747e08f557 missing blob 26f5d7bb9a418d59c6d766388b4320ad26f63035 missing blob 26f67cccc37c3c4dd31164ba3499572ffa937d49

broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree f7a02df4242f09f713b70b1f8e74dbf7f0068566
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree c4ec76f4e8721e4b5b9651dac62c233980fc9d78
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree c16336297d8ad2117b98126f32cc0fdeb42f94eb
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree 7b9962eb29d2b41bd58bb1dd0be529c73bf08012
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree 3a8587c8c755788804b430cc332a1aecebf389d4
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree f0a90cc615da5833d95ff5cb2c96e44d20893de6
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree 0a8763981774041f3fee0a71e016dcaa096fa3f8
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree c1416c92121e10ef4a8d0450133a9e92c90cc9c9
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree 38585f38d977552fbcbc30885c6e02a707adfdeb
broken link from    tree 26f78f3e3a0156b75c84ef01d61130454bc4e673
              to    tree 73cc6db7bf26213d2768966a3bfa03d8ccb9f6cf

[thousand of warnings like this]