[rfc][icedtea7] Handle alternative Kerberos credential cache locations

[Andrew Hughes]

----- Original Message -----
> Hi Andrew,
> 
> On 08/14/2013 10:54 AM, Andrew Hughes wrote:
> > ----- Original Message -----
> >> Hi,
> >>
> >> Kerberos 1.11 introduced a new configuration variable to override the
> >> default location of the credential cache at build time. Fedora 18 and up
> >> have used this new configuration variable to define an alternate default
> >> cache location (/run/user/$UID/krb5cc/tkt). This bug was initially
> >> reported against Fedora [1].
> >>
> >> On Linux and Solaris systems, FileCredentialsCache.getDefaultCacheName()
> >> defaults to the previously hard-coded location (/tmp/krb5cc_$UID). This
> >> location will be incorrect if Kerberos was built with an alternative
> >> credential cache location set. Since this credential cache location can
> >> be arbitrary, we need to query the Kerberos API for the correct
> >> location. This patch implements this query using a new JNI call, which
> >> adds a dependency on libkrb5 for Linux and Solaris systems.
> >>
> >> This patch was prepared against icedtea7-forest/jdk, changeset
> >> afaedb56b499.
> >>
> >> 2013-08-12  Elliott Baron <ebaron at redhat.com>
> >>       * make/sun/security/Makefile: Build krb5/internal/ccache on Linux
> >> and Solaris.
> >>       *
> >> src/share/classes/sun/security/krb5/internal/ccache/FileCredentialsCache.java:
> >> Replace
> >>       hard-coded cache location with native call to Kerberos API.
> >>       * make/sun/security/krb5/internal/ccache/Makefile: New file; builds
> >> JNI wrapper for
> >>       needed Kerberos API.
> >>       *
> >> src/solaris/native/sun/security/krb5/internal/ccache/krb5ccache.c: New
> >> file; JNI function
> >>       to query default cache location from Kerberos API.
> >>
> >> Thanks,
> >> Elliott
> >>
> >> [1] https://bugzilla.redhat.com/show_bug.cgi?id=991170
> >>
> >>
> > I'm building this now and will commit if it builds fine.  I noticed that
> > the
> > copyright header on the new Makefile didn't match that on the new source
> > file,
> > so I copied over the one from the source file, as I assume you didn't
> > change
> > to working for Oracle mid-patch ;)
> 
> Thanks! As far as I know I didn't switch jobs mid-patch ;). The header
> was actually intentional, since this new Makefile is heavily derived
> from one of the existing JNI Makefiles. I was a bit unsure how to deal
> with this header.
> 

I revised them slightly in the committed version:

http://icedtea.classpath.org/hg/icedtea7-forest/jdk/rev/3e4e1a4ef584

Did you actually test on Solaris?  If not, you probably shouldn't enable
it there.  I have a feeling the _GNU_SOURCE won't work there.

> Thanks,
> Elliott
> 

-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07