I've been evaluating JGit 7.2.0 for inclusion in a Jenkins git client
plugin release.
I've found a file handle leak in Pack.java. There were
significant changes in Pack.java between 7.1.0 and 7.2.0 as part
of 51d6c63fe1602a by Martin Fick.
I suspect that the open/close accounting is not always handled
correctly. I've not yet been able to create a repeatable failure
case in a JGit test but I've documented a repeatable failure case
Can you retry this with a local build of jgit stable-7.2 with a revert of
and check if this fixes your issue ?
This change reduced the scope of synchronized sections to reduce contention
between a thread reading/writing a pack and other threads loading indices
which formerly used the same lock object.
Maybe this introduced some race in the way the modified methods are used in
your environment.
I'm posting here in hopes that someone that is more familiar with
the JGit internals could look at the stack traces and suggest
places that I should investigate more fully.
The stack traces are:
I'll continue investigating in any case and will share results
as they develop.
Mark Waite
jgit-dev mailing list
jgit-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/jgit-dev
