Locking bug on Mac OS X over AFP

[E. Wing]

bump

On 2/8/08, E. Wing <ewmailing at gmail.com> wrote:
> Hello, I believe I have found a reproducible locking bug with
> Mercurial (0.9.5). Using two Macs, both running Leopard, when using
> the Apple File Sharing Protocol (AFP), a clone from a client machine
> will permanently lock the original repo, which prevents further
> access.
>
> To reproduce:
>
> On Mac #1, create an initial repository:
> 1) mkdir LOCKBUG
> 2) cd LOCKBUG
> 3) echo "hello" > foo.txt
> 4) hg init
> 5) hg add
> 6) hg commit -m"Initial"
>
> On Mac #2, mount Mac #1 using AFP (done through SHARED discovered
> Bonjour list in Finder) and then run on Mac #2
> 7) hg clone /Volumes/ewing/LOCKBUG LOCKBUG
>
> Repository will now be locked permanently. So for example, trying to
> clone again on Mac #2 yields:
> 8) hg clone /Volumes/ewing/LOCKBUG LOCKBUG2
> abort: Resource busy: /Volumes/ewing/LOCKBUG/.hg/store/lock
>
> And trying to work with the repository on Mac #1 yields:
> 9) echo "world" >> foo.txt
> 10) hg commit -m"2nd"
> waiting for lock on repository /Users/ewing/LOCKBUG held by
> 'mac2.local:1813'
>
> (This process waits forever.)
>
> I am using the Mercurial installation from the OS X installer package
> described as:
> 2008-02-05: Version 0.9.5 + fixes through 2008-02-05
> Mercurial 0.9.5-b7f44f01a632 for OS X 10.5
>
> Both systems have this same version installed. Both computers are
> running Mac OS X 10.5.1.
>
> Thanks,
> Eric
>