On Mon, Dec 18, 2006 at 11:26:10AM +0000, Chavdar Ivanov wrote:
> On 12/18/06, Christos Zoulas <christos@astron.com> wrote:
> >In article <20061218135116.161a6a45@garlique.algebras.org>,
> >George Michaelson  <ggm@apnic.net> wrote:
> >
> >>sure. Here is the kdump output over the ktrace of java -jar
> >> 25114      1 java     CALL  readlink(0x80555e1,0xbfbfb710,0xfff)
> >> 25114      1 java     NAMI  "/emul/linux/proc/self/exe"
> >> 25114      1 java     NAMI  "/emul/linux"
> >> 25114      1 java     NAMI  "/emul/linux/proc/self/exe"
> >> 25114      1 java     RET   readlink 1
> >
> >This is the problem. readlink returns only 1 which means it returns /
> 
> On i386 4.99.5 as of 05/12 readlink returns the normal length and java
> works. On the same 4.99.5 as of 10/12 it returns 1. Something has been
> done in between - perhaps in readlink itself?

FYI, this should be fixed by pooka's first change to vfs_lookup.c yesterday.
(at least, it'll work as well as it did before my changes on dec 9.)

-Chuck