Created attachment 313685 [details] ulogd-1.24-r3.ebuild The changes in this ebuild - relative to 1.23-r2 - are as follows: * Use EAPI-4 with several clean-ups * Bump to git HEAD (just one new patch relative to previous ebuild) * Apply patch to to stop kernel headers being imported from source tree * Use patch for LDFLAGS fix instead of a sed hack * Remove largefile hack. It's already handled by the AC_SYS_LARGEFILE macro. * Simplify runscript modification process * Remove Rules.make sed hack (it's a no-op) * Change econf arg construction to stop features from being automagically enabled Regarding the last point, --without-mysql doesn't actually stop it from building the mysql module. The solution is to specify --with-mysql (if desired) or not to specify it at all. Therefore, I've employed that pattern for all of the optional features. The previous ebuild is applying all patches up to commit 835110. Bizarrely, it does that through a patch named "ulogd-glsa-200805.patch" even though it does not pertain to any such GLSA and the changes encompass more than just security fixes. Because these patches are immensely valuable and because the v1 branch is otherwise dead, I think it is a good idea to patch to the HEAD commit as before. However, instead of using a patch, I've changed the ebuild to use the git-2 eclass and pinned it to the final commit.
Created attachment 313687 [details] files/ulogd-0.98
Created attachment 313689 [details, diff] files/ulogd-1.24-ldflags.patch
Created attachment 313691 [details, diff] files/ulogd-1.24-linux-headers.patch
CC'ing QA as Wormo may not be able to attend to this imminently.
Created attachment 313693 [details] ulogd-1.24-r3.ebuild Two more fixes, inspired by the ulogd-2.0.0_beta4 ebuild: * Install manpage * Use $ROOT variable in post_install()
Is the version number in the summary correct? (In reply to comment #0) > [...] I've changed the ebuild to use the git-2 eclass and pinned it to the > final commit. That's a very bad idea. <http://devmanual.gentoo.org/ebuild-writing/functions/src_unpack/cvs-sources/#disadvantages-of-cvs-sources>
Yes, the new ebuild would be 1.23-r3 so it's correct, though I accidentally wrote "1.23-r4" in the summary when the bug was opened. As for the stated criticisms of using 'CVS', I think they are somewhat overblown but that's fine; I'll have it use a patch instead.
Created attachment 313703 [details] ${DISTDIR}/ulogd-1.24-commit-d2ecdc.incr.patch.xz
Created attachment 313707 [details] ulogd-1.24-r3.ebuild (no git-2 eclass usage) Variant of the ebuild which, like 1.24-r2, uses an incremental patch instead of git to apply upstream post-release fixes. Generation of the patch is painful due to inconsistencies between release tarballs and their equivalent commits (which is why I started using the git eclass in the first place). In any case, the outcome of either ebuild should be the same.
Created attachment 313709 [details] ulogd-HEAD-d2ecdcd.tar.gz (snapshot) Simpler to use a snapshot, I belatedly realize. Amusingly, it turns out to be smaller then the patch.
Created attachment 313713 [details] ulogd-1.24-r3.ebuild New ebuild which uses a snapshot tarball instead of monkey-patching. Tested and working.
Corrected summary as I failed entirely to observe the mistake pointed out by Ulrich. Apologies. Nothing further to add.
This package needs a proxy maintainer for this and libnetfilter_acct: http://www.gentoo.org/proj/en/qa/proxy-maintainers/
Created attachment 342744 [details] ulogd-1.24_p20080618.tar.gz Snapshot of ulogd-1 git[0] at it's latest revision. [0]: https://git.netfilter.org/ulogd/
Created attachment 342746 [details, diff] ulogd-1.24-add-syslog-in-config-example.patch Patch to add ulogd_SYSLOG output plugin's configuration example in ulogd.conf
Created attachment 342750 [details, diff] ulogd-1.24-add-user-option Patch to add `-u` option to run as specified user
Created attachment 342752 [details, diff] ulogd-1.24-fix-include-in-libipulog-Makefile.patch Patch to use $CC not $LD when building extensions. Thanks to Kerin Millar.
Created attachment 342754 [details, diff] ulogd-1.24-fix-plugins-ldflags.patch Patch to use $CC not $LD when building extensions. Thanks to Kerin Millar.
Created attachment 342756 [details, diff] ulogd-1.24-fix-include-in-libipulog-Makefile.patch Patch to prevent include from /usr/src/linux/. See also #285376
Created attachment 342758 [details, diff] ulogd-1.24-sqlite.patch Fixes for sqlite support
Created attachment 342760 [details, diff] ulogd-1.24-suppress-parallel-build-warnings.patch Patch that theoretically allows parallel build and install. In practice, parallel build is still broken due to poorly written Makefile, while parallel install works fine after this patch. It also fixes QA warnings about "jobserver unavailable: using -j1"
Created attachment 342764 [details] ulogd-1.24_p20080618.ebuild Add snapshot ebuild for ulogd-1.24. This snapshot include lots of fixes which are not included in 1.24, but can be found in ulogd-1 git. These fixes are almost completely superseed our GLSA patch, also some things are missing from this patch, but can be found in the git. For this reason, all changes that were in GLSA patch, but not in git are separated in several patches which are applied separately, namely ulogd-1.24-add-user-option.patch, ulogd-1.24-add-syslog-in-config-example.patch and partly ulogd-1.24-sqlite.patch which was extended. Also we include patch from Kerin Miller as well as couple of my own small patches. This ebuild rely on recently added ulogd-2.init and ulogd-2.logrotate files.
Created attachment 342766 [details, diff] ulogd-1.24-add-user-option.patch Patch to add `-u` option to run as specified user
@proxy-maint team if you find ulogd-1.x versions worth keeping in tree please replace current ulogd-1.{23,24} ebuilds with the one suggested above.
After some discussion with maintainer, we agreed to drop this from tree, so closing as WONTFIX. Please, migrate to new ulogd, that is in tree(currently it is 2.0.2)
