Russ Allbery [Sun, 1 May 2005 00:40:06 +0000 (00:40 +0000)]
* Overhaul the module build rules.
- Use debhelper to build the module binary package.
- Let debhelper handle the module postinst and prerm control scripts.
- Properly support modules_config and modules_clean.
- Break the build process into configure, build, install, and binary
rules, paralleling the regular build rules and making it easier to
maintain the make-kpkg build targets.
- Include the kernel version in the long description.
- Update standards version of module packages to 3.6.1.
- Include the copyright and basic documentation in the module binary
package. Otherwise, policy requires an exact versioned dependency
on the package containing the copyright file.
Also note that the logic to turn off optimization when requested doesn't
work as documented in the configure script.
* Update standards version to 3.6.1.
- Install Debian::OpenAFS::ConfigUtils in /usr/share, not /usr/lib.
- Always enable debugging symbols and disable optimization if noopt is
set in DEB_BUILD_OPTIONS.
- /etc/init.d/openafs-fileserver restart starts bosserver if it's not
already running.
* README.Debian: Document that the client cache partition must be ext2
or ext3 and that XFS and ReiserFS will not work. upserver and
upclient are now provided. Provide some information about why
kaserver is not provided. (Closes: #249315)
Also clean up the formatting of README.Debian a little.
* Put the compiled modules into KPKG_DEST_DIR if set. This appears to
be preferred over DEB_DEST by the make-kpkg documentation. Thanks,
Jonas Smedegaard. (Closes: #206403)
Sam Hartman [Mon, 4 Apr 2005 04:49:58 +0000 (04:49 +0000)]
Rerun regen.sh without autoconf2.13 installed.
Apparently the debian magic to choose the right version of autoconf interacts badly with aclocal
and afs
Sam Hartman [Sun, 9 Jan 2005 01:03:14 +0000 (01:03 +0000)]
* New upstream version
* Target for unstable not experimental
* New CellServDB
* OpenAFS seems not to have a changelog available, so stop shipping it.
* Build openafs.ko not libafs.ko so the right name appears in lsmod
Sam Hartman [Sun, 30 May 2004 21:12:55 +0000 (21:12 +0000)]
* use make only_libafs not dest_only_libafs
- AS a result, build times are much faster, Closes: #144781
* Correct module build scripts to install module package from new location
* Require 1.3.64 clients for modules built by 1.3.64
* installo penafs.o not libafs-*.o
Sam Hartman [Sun, 30 May 2004 20:19:51 +0000 (20:19 +0000)]
linux-test3.m4: The previous hack didn't work because forcing
modversions to yes breaks 2.6 handling of export table. Instead,
force the test for what modules to build to only build one module.
Sam Hartman [Sun, 30 May 2004 19:24:21 +0000 (19:24 +0000)]
linux-test3.m4: Always pretend we are using modversions so we only build
the kernel module that corresponds to our sp/mp configuration. Building both for Debian is wrong.
This is a hack.
=== below are targets to be committed === M
/home/hartmans/debian/openafs/openafs/src/cf/linux-test3.m4
Sam Hartman [Sun, 30 May 2004 19:07:46 +0000 (19:07 +0000)]
in osi_module.c:
Don't set sys_settimeofday to 0 because
it doesn't seem like the symbol is defined or used on the codepath where it is set.
This is a hack; upstream solution is preferred.
Sam Hartman [Sun, 30 May 2004 19:03:22 +0000 (19:03 +0000)]
* Apply patch from
/afs/cs.cmu.edu/user/jhutz/junk/openafs-patches/openafs-build-on-linux26
which correctly generates 2.6 modules.
* Make sure makefile generation script is executable in debian/rules
Sam Hartman [Fri, 21 May 2004 04:33:38 +0000 (04:33 +0000)]
* Delete src/venus/lock.h as it is deleted upstream
* src/pam/Makefile.in should not link in libs; it uses pic objects
* Fix merge errors in src/sys and src/ubik