threads / discuss / 11041

ok for git to delete temporary packs on write error?

Subject: ok for git to delete temporary packs on write error?

## tl;dr

3 messages between Nov 28, 2007 and Nov 28, 2007.

replies: 2people: 2as markdown or json

David Tweed· Nov 28, 2007, 07:42 UTC · lore

Hi, I'd like to check if there's any reason in the overall design of git which would make deleting tmp_pack's that have suffered write errors a bad idea? (Before I look further into this I may be missing a good reason why they shouldn't be auto-deleted.)

My encounter with this comes from using an almost full usbstick which I discovered when I was poking around for other reasons several partial packs from occasions (separated by weeks) where gc failed. On each failure I'd removed stuff from the drive to clear space and done a successful gc but hadn't thought to check below .git for removable stuff so they'd just accumulated.

Below is a output of a test session:

$ git version git version 1.5.3.6

$ git gc --aggressive --prune
Generating pack...
Done counting 22216 objects.
Deltifying 22216 objects...
 100% (22216/22216) done
Writing 22216 objects...
fatal: sha1 file '/media/usbdiskc/v.git/objects/tmp_pack_QCYYAi' write
error (No space left on device)
error: failed to run repack

$ ls -l /media/usbdiskc/v.git/objects/ total 3944 drwxr-xr-x 2 sis05dst sis05dst 2048 2007-11-28 07:25 info drwxr-xr-x 2 sis05dst sis05dst 2048 2007-11-28 07:25 pack -rwxr-xr-x 1 sis05dst sis05dst 4034560 2007-11-28 07:25 tmp_pack_QCYYAi -rw------- 1 sis05dst sis05dst 0 2007-04-18 23:02 tmp_pack_RYLguI

-- 
cheers, dave tweed__________________________
david.tweed@gmail.com
Rm 124, School of Systems Engineering, University of Reading.
"we had no idea that when we added templates we were adding a Turing-
complete compile-time language." -- C++ standardisation committee

Nicolas Pitre· Nov 28, 2007, 15:01 UTC · re: David Tweed · lore

Re: ok for git to delete temporary packs on write error?

On Wed, 28 Nov 2007, David Tweed wrote:

Show 12 quoted lines

> Hi, I'd like to check if there's any reason in the overall design of
> git which would make deleting tmp_pack's that have suffered
> write errors a bad idea? (Before I look further into this I may be missing
> a good reason why they shouldn't be auto-deleted.)
> 
> My encounter with this comes from using an almost full
> usbstick which I discovered when I was poking around
> for other reasons several partial packs from occasions
> (separated by weeks) where gc failed. On each failure
> I'd removed stuff from the drive to clear space and done
> a successful gc but hadn't thought to
> check below .git for removable stuff so they'd just accumulated.

I'd suggest they get removed a part of the prune command.

Nicolas

David Tweed· Nov 28, 2007, 15:48 UTC · re: Nicolas Pitre · lore

Re: ok for git to delete temporary packs on write error?

On Nov 28, 2007 3:01 PM, Nicolas Pitre <nico@cam.org> wrote:

Show 6 quoted lines

> On Wed, 28 Nov 2007, David Tweed wrote:
>
> > Hi, I'd like to check if there's any reason in the overall design of
> > git which would make deleting tmp_pack's that have suffered
> > write errors a bad idea? (Before I look further into this I may be missing
> > a good reason why they shouldn't be auto-deleted.)

[snip]

> I'd suggest they get removed a part of the prune command.

Mmm, I hadn't looked at the source and spotted the prune is actually a separate process to the repack. From a policy POV is it ok to assume any tmp_pack*'s existing when prune is run can be deleted? (I know you're not supposed to be having any other git ops going on whilst git gc works.)

Now I know they aren't being kept around for, say, "emergency error recovery" I may look to see if I can come up with an acceptable patch. (Don't hold your breath waiting.)

-- 
cheers, dave tweed__________________________
david.tweed@gmail.com
Rm 124, School of Systems Engineering, University of Reading.
"we had no idea that when we added templates we were adding a Turing-
complete compile-time language." -- C++ standardisation committee