Super, thanks.
I see from your JIRA issue that somebody else stumbled upon the same
problem a few hours after we diagnosed this last night! :)
Regards,
Chris
On 03/27/2014 07:46 PM, Mark Waite wrote:
Thanks for detecting and diagnosing that problem. The JGit 3.3.0
upgrade happened in git-client-plugin 1.6.5. It is also in 1.6.6. The
next release of git-client-plugin (1.7.0) will include the fix.
Hopefully, I'll also have time to write a test which specifically
asserts that bug "stays fixed".
I submitted https://issues.jenkins-ci.org/browse/JENKINS-22391 to track
the bug.
Thanks,
Mark Waite
On Thu, Mar 27, 2014 at 12:20 PM, Christopher Orr <[email protected]
<mailto:[email protected]>> wrote:
Hi all,
We discovered a subtle bug yesterday with the Git Client plugin and
its use of jgit (even if you're using the default command-line git
implementation).
I just noticed that this bug was probably fixed today by a pull
request, but it's worth letting Jenkins users know about this.
If Jenkins clones a git repo on a machine running Java 1.6, the repo
will be created with "core.symlinks = false" in the .git/config file.
This causes Bad Things to happen if your git repo happens to rely on
the use of symlinks.
This happens because the Git Client plugin no longer uses "git
clone" to clone a repo, but rather a combination of "git init" then
"git fetch", and because the "git init" behaviour is implemented by
jgit, even if you're using the default git CLI implementation.
As of jgit 3.3.0, if you're running on a JVM which does not support
symlinks (Java 1.6 or older), then the library will explicitly set
"core.symlinks" to false when creating the .git directory.
This means "git init" will create a repo which claims *not* to
support symlinks, even if your file system *does* support symlinks.
"git fetch" will then be executed (via the git binary, rather than
jgit) and will respect the erroneous "core.symlinks" config value.
Therefore you end up with files rather than the symlinks you
expected in your cloned repo.
(We actually saw this on an OS X 10.9 server with Jenkins running
under Java 1.7 (with Java 1.6 also on the server as the non-default
Java install), but jgit -- for some reason -- seems to think it's
running on Java 1.6 and so believes symlinks aren't supported).
Partly this seems to be a Git Client problem, but also (arguably) a
jgit problem, as they explicitly configure new repos to reject
symlinks, under the assumption that the repo will always be
manipulated via jgit.
But at least this merge from today would appear to fix the Git
Client plugin by removing the use of jgit for "git init" for the CLI
implementation:
https://github.com/jenkinsci/__git-client-plugin/commit/__f5b27a9
<https://github.com/jenkinsci/git-client-plugin/commit/f5b27a9>
So, If you see any weird problems with git repos and symlinks, you
know why! :)
Regards,
Chris
--
You received this message because you are subscribed to the Google
Groups "Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to jenkinsci-dev+unsubscribe@__googlegroups.com
<mailto:jenkinsci-dev%[email protected]>.
For more options, visit https://groups.google.com/d/__optout
<https://groups.google.com/d/optout>.
--
Thanks!
Mark Waite
--
You received this message because you are subscribed to the Google
Groups "Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected]
<mailto:[email protected]>.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups "Jenkins
Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
