OpenAFS strips everything except for the fileserver and volserver, so
for right now only provide debugging information for those two binaries.
Document the new debugging support in README.Debian.
Russ Allbery [Thu, 15 Jun 2006 14:21:48 +0000 (14:21 +0000)]
r2165@windlord: eagle | 2006-06-07 12:56:09 -0700
* Use a better method of optionally running make distclean.
* Add build-arch and build-indep targets, just in case.
Russ Allbery [Thu, 15 Jun 2006 14:21:36 +0000 (14:21 +0000)]
r2164@windlord: eagle | 2006-06-07 12:52:30 -0700
* Regenerate the man pages on build since the pre-generated ones were
done with a very old version of pod2man.
Russ Allbery [Thu, 15 Jun 2006 14:21:13 +0000 (14:21 +0000)]
r2162@windlord: eagle | 2006-06-07 12:09:03 -0700
* Update debconf templates to follow the best practice recommendations
from the Developer's Reference.
Russ Allbery [Thu, 15 Jun 2006 14:20:59 +0000 (14:20 +0000)]
r2161@windlord: eagle | 2006-06-07 12:01:55 -0700
* When creating a new file server in afs-newcell, use a set of tuning
flags from Harald Barth rather than relying on the defaults. This
doesn't help with general bos create; long term, the defaults need to
be fixed.
* Add a note in README.servers about fileserver tuning options.
Russ Allbery [Thu, 15 Jun 2006 14:20:34 +0000 (14:20 +0000)]
r2159@windlord: eagle | 2006-06-07 11:27:08 -0700
* Include all of AFS's local XDR implementation in the build for
shlibafsrpc, fixing problems with undefined symbols in the PAM module
on AMD64.
* Exit successfully in the openafs-client init script if no module is
installed, since otherwise openafs-client will fail to install until
one builds a module. This restores previous accidental behavior due
to a bug in debhelper. (Closes: #362695)
Russ Allbery [Wed, 29 Mar 2006 20:56:10 +0000 (20:56 +0000)]
* Since 2.6 is the supported kernel for etch, build the userspace with
a sysname of *_linux26 where supported. This only affects the sys
command. (Using livesys instead is better.) (Closes: #358118)
Russ Allbery [Wed, 29 Mar 2006 04:29:43 +0000 (04:29 +0000)]
* Subsume openafs-krb5 into this source package.
- Acknowledge NMU. (Closes: #304933)
- Update to the OpenAFS aklog and asetkey. (Closes: #196094)
- Backport asetkey from upstream CVS.
- Include ka-forwarder from the old migration toolkit.
Russ Allbery [Thu, 9 Feb 2006 04:51:23 +0000 (04:51 +0000)]
* Document the need to restart ptserver and vlserver on the existing
database servers when adding a new one. Thanks, Niklas Edmundsson.
(Closes: #349659)
Russ Allbery [Tue, 27 Dec 2005 00:23:42 +0000 (00:23 +0000)]
* Call kernel-version correctly from prep-modules, which fixes build
issues with manually-built kernels. (Closes: #344851)
Also always trust the KVERS determined by the Makefile in prep-modules
if available, since otherwise we end up with broken packages when the
makefile installs the module assuming one package name and the control
file tries to build a package with a different name.
Clean up a few remaining remnants of @sys handling when building kernel
modules. Since the kernel module build never runs any sort of install
target, this shouldn't be needed.
Russ Allbery [Thu, 22 Dec 2005 02:54:56 +0000 (02:54 +0000)]
* Apply upstream patch from Chas Williams to be more careful about what
dentry is used in renames, avoiding a kernel fault when renaming
dentries accessed through multiple paths (such as RO and RW).
Russ Allbery [Wed, 21 Dec 2005 20:00:05 +0000 (20:00 +0000)]
* Include volinfo, not voldump in openafs-fileserver.
* Use make install into debian/tmp rather than make dest. make dest is
mostly for Transarc compatibility and make install looks more like a
typical package and makes it easier to evaluate differences in FHS
location between upstream and Debian. As a side effect, there's no
need to mess with the @sys link. (Closes: #165909)
* Install more files with dh_install rather than install commands in
debian/rules so that *.install shows a more complete picture of what
goes into each package.
* Don't ignore make distclean errors.
Russ Allbery [Wed, 21 Dec 2005 04:17:23 +0000 (04:17 +0000)]
* Stop unconditionally forcing the compiler to cc or gcc in configure,
allowing, among other things, building the kernel module with a
compiler other than the default system compiler. (Closes: #338306)
Russ Allbery [Wed, 7 Dec 2005 06:39:00 +0000 (06:39 +0000)]
* openafs-client should only recommend the AFS kernel module. Many of
the client programs will work in unauthenticated mode without it, and
openafs-fileserver depends on it (but doesn't require the kernel
module). (Closes: #342095)
Russ Allbery [Wed, 7 Dec 2005 05:10:27 +0000 (05:10 +0000)]
* Install more of the standard OpenAFS utilities. (Closes: #138851)
- openafs-dbserver: Install prdb_check and vldb_check.
- openafs-fileserver: Install voldump.
- openafs-client: Install fstrace, scout, and up (as afs-up).
* Do not install kpwvalid. The default version does little useful and
the kpwvalid binary must be in AFS for it to take effect.
Also remove an old rm from debian/rules that no longer does anything.
Russ Allbery [Wed, 7 Dec 2005 04:03:15 +0000 (04:03 +0000)]
* Overhaul the way kernel modules are built.
- Recommend linux-image rather than kernel-image for newer kernels.
- Always recommend the kernel package; since it's only a recommends,
there's no need to special-case the hand-built kernel case.
- Be less fragile about how the kernel version is used in prep-modules
to support other ways kernel modules are built. (Closes: #341836)
- Add Build-Depends and Uploaders to the openafs-modules-source
control file.
- Remove unused parts of prep-modules and heavily comment it.
Russ Allbery [Thu, 6 Oct 2005 23:01:24 +0000 (23:01 +0000)]
* Install afsd in /sbin rather than /usr/sbin. While putting /usr in
AFS isn't as common as it once was, no reason to make it harder than
it needs to be.
* libpam-openafs-kaserver must depend on openafs-client, not just
recommend it, since acquiring tokens requires that the client be
running.
* Expand README.Debian and include information about PAM configuration.
* Link libpam-openafs-kaserver's doc directory to openafs-client.
* Remove the init script check of the cache file system and instead
improve the check in afsd. The init script check used stat, which is
in /usr/bin and therefore not fair game for init scripts.
* When kill_all_afs is enabled in /etc/openafs/afs.conf, only kill AFS
processes if the runlevel is 0 or 6. You really only want this when
the system is shutting down. (Closes: #325347)
Russ Allbery [Fri, 26 Aug 2005 05:19:41 +0000 (05:19 +0000)]
Clean up error handling and reporting a little bit more. Add a warning
before creating the pt database, saying to expect a few harmless errors.
Add an unwind for deleting the root.afs volume, in case starting the
client fails.