If this is indeed issue 769, I suspect it will be fixed by the 2.3 upgrade (since they upgraded JGit to 1.1).

We can test again then.

And for the record, it doesn't *crash* gerrit. It hits an exception that's logged and it fails for the user, but it doesn't *crash*

(In reply to comment #2)

And for the record, it doesn't *crash* gerrit. It hits an exception that's
logged and it fails for the user, but it doesn't *crash*

Thanks for investigating this in the logs, and for reporting!

Works for me with the 2.3 upgrade:

$ git clone --depth 1 https://gerrit.wikimedia.org/r/p/mediawiki/core.git test-shallow-clone
Cloning into 'test-shallow-clone'...
remote: Counting objects: 31773, done
remote: Finding sources: 100% (31773/31773)
remote: Total 31773 (delta 12783), reused 16435 (delta 12783)
Receiving objects: 100% (31773/31773), 291.11 MiB | 1.82 MiB/s, done.
Resolving deltas: 100% (12783/12783), done.

Marking FIXED.

Hmmmm, strange is my following observation (each clone into a fresh directory):

git clone ssh://<username>@gerrit.wikimedia.org:29418/mediawiki/core.git
du -h .

> 178 MB

git clone --depth 1 ssh://<username>@gerrit.wikimedia.org:29418/mediawiki/core.git
du -h .

> 360 MB [sic]

> Shallow clone needs 360 MB ??? I am puzzled.

some more examples, where shallow clones are smaller:

git clone git://gitorious.org/owncloud/owncloud.git ==> 39M

git clone --depth 1 git://gitorious.org/owncloud/owncloud.git ==> 34M

git clone https://github.com/Pita/etherpad-lite.git ==> 4,8M

git clone --depth 1 https://github.com/Pita/etherpad-lite.git ==> 2,6M

for info only (2012-05-19)
+ git clone --depth 1 https://gerrit.wikimedia.org/r/p/mediawiki/core.git ==> 84M
+ git clone https://gerrit.wikimedia.org/r/p/mediawiki/core.git ==> 184M