- Fri Oct 29 2010 Thierry Vignaud <tv@mandriva.org> 1:4.8.1-1mnb2
+ Revision: 590113
- new release: 4.8.1
- actually drop 4.6.0 tarball
- bump release so that cooker's rpm is newer than 2009.1's one
- fix build on x86_64
- replace %mkrel with %manbo_mkrel for Manbo Core 1
- replace %{_datadir}/man by %{_mandir}!
- replace %_datadir/man by %_mandir!
- fix URL
+ Christophe Fergeau
- Initial import of rpm 4.8.0
* rpm-4.6.0.tar.bz2
* rpm-4.8.0.tar.bz2
new rpm release
* rpm-4.6.0-rc1-revert-unused-skipDir-functionality.patch
* rpm-russian-translation.patch
* rpm-4.6.0-rc1-compat-PayloadIsLzma.patch
* rpm-4.6.0-lua-add-sources-and-patches-only-once.patch
* rpm-fix-corequisites.patch
* rpm-fix-islegacyprereq.patch:
* rpm-4.6.0-rc1-drop-skipping-ldconfig-hack.patch
drop all these patches which were merged upstream
* rpm-4.6.0-rc1-speedup-by-not-checking-same-files-with-different-paths-through-symlink.patch
rediffing wasn't trivial because some functions used by this patch were dropped upstream so it's disabled for now
* rpm-4.6.0-rc3-xz-support.patch
xz support was integrated upstream, only a few diffs remain between our implementation and ours
* rpm-4.6.0-rc1-python-writeHdlist.patch
drop patch, it needs to be reworked because librpm python bindings changed, and it no longer seems to be used
* mips_define_isa_macros.patch
* rpm-4.6.0-rc1-protect-against-non-robust-futex.patch
* rpm-4.4.2.3-rc1-transmeta-crusoe-is-686.patch
* rpm-check-file-trim-double-slash-in-buildroot.patch
* rpm-4.6.0-rc1-bb-shortcircuit.patch
* rpm-4.6.0-rc3-no_rm_-rf_DOCDIR.patch
* rpm-4.6.0-rc1-syslog.patch
* rpm-4.4.2.3-rc1-fix-debugedit.patch
* fix_stack_protector_check.patch
* rpm-4.6.0-rc1-buildlang.patch
* rpm-4.4.2.2-gendiff-improved.patch
* rpm_arm_mips_isa_macros.patch
* rpm-4.6.0-rc1-fix-nss-detection.patch
* rpm-4.6.0-rc1-weakdeps.patch
* rpm-4.4.2.2-rpmqv-ghost.patch
* rpm-forbid-badly-commented-define-in-spec.patch
* rpm-allow-conflicting-ghost-files.patch
* introduce-_patch-and-allow-easy-override-when-the-p.patch
* rpm-4.6.0-rc1-non-pre-scripts-dont-fail.patch
* rpm-4.6.0-rc1-copyright-tag.patch
* rpm-4.6.0-rc1-serial-tag.patch
* mips_macros.patch
* rpm-4.6.0-rpm5-tags.patch
* rpm_add_armv5tl.patch
rediff patches
* rpm-rpmbuild-check-useless-tags-in-non-existant-binary-packages.patch
* rpm-4.6.0-rc1-provides-obsoleted.patch
* rpm-4.6.0-rc2-manbo-rpmrc-rpmpopt.patch
* introduce-apply_patches-and-lua-var-patches_num.patch
* rpm-4.6.0-no-doc-conflicts.patch
* rpm-4.6.0-rc1-filetriggers.patch
* rpm-4.6.0-rc1-read-macros_d-dot-macros.patch
* introduce-_after_setup-which-is-called-after-setup.patch
rediff patches and adjust to rpm 4.8.0 API changes
* rpm-4.6.0-ordererase.patch
disabled for now since it breaks urpmi test suite
* rpm-4.6.0-rc1-headerIconv.patch
rediff and fix compile warning
- create temporary branch to finish rpm 4.8 integration
- don't diverge from upstream wrt EVR comparisons
- add missing calls to rpmluaPop to patch 159, might fix #50579
- Revert switch to db4.7
- Update to rpm 4.6.0
+ Funda Wang
- rebuild for libpopt file path change
- temporarily disable graphviz BR
+ Per Øyvind Karlsen
- fix broken bitwise operator usage (thx to anssi for noticing)
- do a new fix for #55810 that doesn't cause same regressions as the previously
revert fix and hopefully no other regressions either.. ;)
(P1010, http://rpm5.org/community/rpm-devel/4011.html)
- revert previous change as it'll break == dependencies on version only
(which even requires-on-release policy requires) :/
- don't skip release comparision when checking conflicts if release value is
missing (P1011 from Jeff Johnson, fixes #55810)
- Avoid adding Lua sources/patches twice when recursing (P1010, git backport)
- Expose packagecolor tag and add new tags from rpm5 as rpm otherwise will break
when these unknown tags might be found in the rpmdb. Notice that this will only
make rpm recognize these, not implement actual support for them.. (P1008)
- build against bdb 4.7
- build against db4.7 rather than db4.6 from now on
- don't wipe out $DOCDIR when using %doc as it will wipe out any files that would
happen to be installed during %install. (P1008)
- fix internal enum value for xz (noticed by jnovy, thx!)
- oops, fix a minor glitch with a final api change
- add versioned requires on liblzma-devel since we *really* need latest api
changes :)
- update lzma payload support to add support for new xz format in parallel(P1007)
- rebuild against liblzma which I had to bump the major of yet again
- actually remember to bump popt release as well this time (popt should REALLY
be split out!!!)
- fix build with new liblzma (updates P1001)
- fix version typo
- switch to use liblzma for lzma payload (P1000, partly derived from rpm5.org & OpenSuSE)
- use LZMA_Alone file format, even with newer versions (P131, from cvs)
- be more consistent and use unlzma in stead of lzma -d -c in rpm2cpio.sh (lzma patch)
- drop unused leftover macros for lzma patch..
- final fix of lzma payload patch, it can now be used properly same way as gzip and bzip2
- correctify lzma fix patch
- switch to patch from cvs for lzma fixes
- update arguments for lzma utils as current old arguments breaks it
- I suck, don't move man pages around at all and stop messing:p
- move man page for perl module to perl-RPM package
- add requires on lzma
- use %{ix86} & %{sunsparc} macros for x86 & sparc
- use %{sunsparc} macro to ensure sparcv9 gets included too
+ Michael Scherer
- fix unowned directory, reported by bitshuffler on irc, causing problem
when installing rpm and using the wrong umask
- rebuild for new python
+ Olivier Blin
- remove n32 support for now (not actually used)
- handle n32 ABI in find-requires and find-provides (from Arnaud Patard)
- switch to armv5tl (from Arnaud Patard)
There are little chances that we'll have armv4tl. Doesn't use
armv5tel as some SoC have broken instructions in the "e" extension
- fix %_isa rpm macros for ARM (from Arnaud Patard)
- mips build fixes: define __isa* macros otherwise rpm doesn't work
(from Arnaud Patard)
- fix stack-protector check: -fstack-protector doesn't error if it's
not supported but issues a warning, building test case with -Werror
the test case fixes the issue (from Arnaud Patard)
- add %mips macro define all mips abi (from Arnaud Patard)
- do not expand _install_langs macro in default /etc/rpm/macros
- fix typo in default /etc/rpm/macros
- explicitely require neon-devel 0.26 (does not build with neon 0.27)
+ Anssi Hannula
- fix ignored Requires(pre) and (post) when they have a plain Requires
counterpart (rpm-fix-corequisites.patch backported from upstream 4.7.1)
- fix legacy prereq detection (rpm-fix-islegacyprereq.patch backported
from upstream 4.7.1)
- map prereq to Requires(pre,preun) instead of nothingness when building
(rpm-map-prereq.patch backported from upstream 4.7.1)
- fix filetriggers to be called on package removal as well (regression
introduced in 4.6.0 package because of an error in rediffing of
filetriggers.patch; fixes bug #52333)
- allow conflicting ghost file types as sometimes the same ghost file
is a file in one package and a symlink in another package (modifies
rpm-4.4.2.2-allow-conflicting-ghost-files.patch)
- require arch-specific libpopt and libpopt-devel
- add back dropped librpm-devel and libpopt-devel provides
+ Arnaud Patard
- define apply_patches if not already done. This fix bootstrapping issues
(eg building current rpm on older rpm or building rpm on a system without
rpm)
- fix dependency on patch as rpm-build is now using "patch -U" which has
been introduced in version 2.5.9-7mdv2009.1.
+ Pixel
- add %apply_patches: it can be used to replace all "%patchN" lines,
but it forces all patches to be "-p1".
- call %_patch and %_after_setup for use in rpm-mandriva-setup-build
(useful for "--with git_repository")
- replace nss-inithack patch with upstream patch
(which is more complete, esp fixes rpmresign in perl-RPM4)
- add patch fixing rpmdsMerge on rpmdsSingle (fixes 07dep.t in perl-RPM4)
- fix segfault triggered by perl-RPM4 tests
- add conflict on urpmi-recover since rpm 4.6.0 dropped support for --repackage
- add patch153: "fix" segfault in mdkapplet (#46323)
- add patch154: fix compilation with Werror=format-security
- merge patch91 (check-dupl-files) into patch111 (check-files fix for "//" in buildroot)
- enable sqlite (was disabled after rpm-4.6.0 switch)
- 4.6.0-rc3
- drop patch153 (buildroot-subpackage), fixed upstream
- rediff patch71, patch83, patch91, patch140, patch145, patch147, patch1005
- fix patch2002 (python-writeHdlist), was broken with rpm 4.6 API
- use %configure (so that libdir is correctly on x86_64)
- 4.6.0-rc2
- add patch152: fix build
- add patch153: buildroot with subpackages issue
- drop already unused %buildnptl (obsolete since rpm uses system libdb)
- drop patch134 (defaultbuildroot), so %defaultbuildroot is no more used,
global %buildroot is used, and BuildRoot is no more handled
- drop patch56 (ppc32). hopefully now unneeded
- drop patch44 (amd64-x86_64 compat). hopefully now unneeded
- drop patch112 (dont-use-rpmio-to-read-file-for-script), only needed on rpm 4.4.8?
- drop patch1002 (default-topdir--usr-src-rpm). it is now /root/rpmbuild
- drop patch1003: drop support for "suggests" the way rpm >= 4.4.7 do it
(all Mandriva packages uses the new "suggests" nowadays)
- drop patch139 (do-not-allow-fileconflict-between-non-colored-file),
since partially applied upstream (47c85270631de173d873b98bc79727e2db203007)
- drop popt (it is no more internal), drop patch1004, patch2004
- drop patch3 ("file" is no more internal)
- drop patch149 (upstream has already got rid of internal db)
- drop patch1450 (problem fixed upstream, though differently)
- drop patch132 (extcond), patch133 has been simplified (butchered) to work without it
- drop patch1133 (integrated in patch133 (weakdeps))
- drop patch100 (rpmbuild-missing), no more needed since we have rpmb_deprecated
- drop patch142, patch143, patch144, patch150 applied upstream
- simplify patch1001 since most is upstream, create patch1006 containing the
PayloadIsLzma compat issue
- adapt patch137 (headerIconv)
- adapt patch31 (syslog)
- adapt/simplify patch114: only specifically read macros.d/*.macros
- adapt patch2000, patch2001 (serial/copyright tag)
- adapt patch133 (weakdeps), drop --suggests/--recommends difference
- rediff patch49, patch64, patch70, patch141, patch146, patch147, patch148,
patch151, patch2002, patch2005
- failing triggers must not block an update (otherwise both packages are kept)
- protect-against-non-robust-futex patch: remove the ugly error messages
when non-root
- ensure stale futex locks do not block (#41868)
- fix broken cpio for hardlink on softlink (#43737)
- remove /usr/lib/libpopt.so.0 symlink (unneeded)
- use external libdb-4.6.so instead of internal one (was DB 4.3.27: December 22, 2004).
(hopefully fixing #41868)
- add patch148 which ensures chroot errors are not ignored
(the bug occured on mandriva build system, the files
were installed on the non rooted system)
- fix memory leak in patch124 (regression introduced on 2008-06-23)
- fix detecting wether filetriggers are activated (through %_filetriggers_dir)
- fix --testing errors introduced by filetriggers
- nicely handle /etc/rpm/macros.cdb disappearance (esp. for people having %_dbapi set there)
- enhance patch124 to use "Dirnames" db before using "Packages" db
(hence much less db accesses in the pathological "COPYING" case)
- rpm must feature PayloadIsLzma = 4.4.2-1 to be the most compatible (?!)
(with SuSE, and with temporary cooker packages with PayloadIsLzma <= 4.4.2.2-1)
- PayloadIsLzma version must be at least 4.4.6-1 to be compatible with mdv2008.0
- fix segfault when transaction fails
- drop our old macros.cdb, which was forcing old values,
esp "verify" which forcing rpm to verify db after each changes
(which was making rpm quite slow when rpmdb was not in cache)
- cleanup: drop not applied global-RPMLOCK patch
- fix filetriggers (librpm was exiting in case of sigpipe in filetrigger script)
- checks on "non packaged binary packages" now depend on %_missing_subpackage_terminate_build (patch147)
- checks on "non packaged binary packages" only done on "mdv" packages (patch147)
- rpmbuild: add patch to ensure some parse errors are really fatal as they should
- add new fatal errors (during package build):
o disallow scriptlets for non packaged binary packages
o "%files foo" for subpackages is now mandatory
- do not call ldconfig in %post/%postun, it is now handled by filetriggers
- replace "%post -p ldconfig" hack with a full featured generic solution:
filetriggers (cf http://wiki.mandriva.com/en/Rpm_filetriggers)
- "#%define foo bar" is now a fatal error
- handle "%posttrans -p ..." with no script body (eg: %posttrans -p /sbin/ldconfig)
- add "requires tar" in rpm-build
- fix build of debugedit (patch143)
- 4.4.2.3
- add patch142 which ensures chroot errors are not ignored
(the bug occured on mandriva build system, the files from package binutils
were installed on the non rooted system)
- drop patches applied upstream: 136, 138
- rediff patch17
- postpone and group ldconfig %post in %posttrans instead of wrongly skipping ldconfig
- rebuild with fixed libneon0.26-devel
- add support for "suggests" and "enhances" in rpmds (needed for "sophie") (nanardon)
- use %_vendor instead of %vendor (thanks-to/required-by nanardon)
- fix typo in recent commit
- fix broken symlink (due to %rpmversion != %srcver for rc1)
- trivial fix for Russian translation (#38713)
- fix file conflicts wrongly allowed on x86_64
- when dropping requires during tsort, do not display "PreReq" when it really is "Requires"
- rpm-mandriva-setup and rpm-mandriva-setup-build requires are only on Mandriva
- add buildlang patch which unsets locales when building packages (for Turbolinux)
- use rpmrc and rpmpopt from rpm-manbo-setup
- get best lang from rpm HEADERI18NTABLE, instead of getting first fuzzy match
(eg: zh_TW matches zh_CN whereas zh_TW entry is available)
- debugedit: remove a wrong check in case %_topdir is /RPM (from Turbolinux)
- add japanese popt translations
- add Turbolinux specific patches:
o re-allow "Serial" and "Copyright" (aliases of "Epoch" and "License")
(to ease Turbolinux migration)
o add writeHeaderListTofile function into python binding
(needed by Turbolinux buildman build system)
- 4.4.2.3 adds ix86 macros files to x86_64
- new release 4.4.2.3-rc1
- libpopt needs a release number different from rpm so that we can reset rpm's
release and increase libpopt's
- drop applied patches: patch82 (ordering), patch98 (bad pointer),
patch108 (dgettext), patch116 (qv), patch119 (without-O2),
patch121 (timeout 60secs), patch127 (rpmbuild --quiet),
patch128 (rpm -K segfault), patch136 (macro on last line),
patch1006 (triggerprein), patch1007 (russian),
patch1008 (display Requires found),
- drop patch135 (truncated rpmProblemString): no more needed, workarounded
upstream
- fix rpmbuild not printing Requires after build (fix backported from 4.4.8) (#36672)
- allow doc conflict in same transaction (#37040)
(it was already allowed in different transactions)
- fix russian translation (#36974)
- %triggerprein were missing in 4.4.2.2, adding them
- fix multiline macro handling on last line of spec file (#27417)
- fix truncated "file conflict" error message in russian (#31680)
- re-introduce temporarily BuildRoot so that rpm builds
- drop patch for SOURCEPACKAGE (unneeded in rpm 4.4.2.2 which keeps compatibility)
- fix defaultbuildroot patch (fix building with %defaultbuildroot and subpackages)
- allow conflicting %ghost files (as used to be in rpm >= 4.4.6)
- modify patch no-doc-conflicts instead of patching it
- for compatibility with rpm 4.4.8, allow conflicting doc files in
/usr/share/man, /usr/share/gtk-doc/html /usr/share/gnome/html
- patch134 introduces %defaultbuildroot to use instead of %buildroot in our global macros
- keep libpopt.so versioning from 4.4.8 (to avoid warnings)
- fix URL
- switch to 4.4.2.2 (using epoch: 1)
- add patch132 and patch133 to handle "Suggests" via RPMTAG_SUGGESTSNAME
- add patches to be compatible with >= 4.4.8 :
o patch1000: handle %buildroot macro
o patch1001: handle suggests via RPMTAG_REQUIRENAME + RPMSENSE_MISSINGOK
o patch1001: lzma-support (integrates patch130 (lzma-fixes), patch131 (lzma_alone))
o patch1002: default %_topdir is /usr/src/rpm
o patch1003: handle RPMSENSE_MISSINGOK (integrates patch129 (do-not-cache-unsatisfied-suggest))
- drop unneeded patches:
o applied: patch63 (applied in lib/package.c), patch68, patch72,
patch77 (%_srcdefattr), patch92 (newtar), patch94 (rpmv3-support),
patch113 (%_docdir_fmt), patch123 (find-lang omf)
o patch1 (no builtin zlib)
o patch2 (rpm is dynamically linked by default)
o patch85 (no more perl module)
o patch87 (was disabling dirname require introduced in 4.4.6)
o patch89 (>= 4.4.6 specific compilation issue)
o patch107 (fixed differently in changeset 6799:446988cfb9c1)
o 4.4.8 specific: patch109, patch117 (dont-replace-config-not-in-db),
patch118 (lowercase platform), patch120 (segfault fix), patch122 (fix
crash with buggy rpm and FILELINKTOS), patch125 (popt.h fix), patch126
(platform32 support)
o 4.4.8 specific?: patch115 (dont-clean-buildroot-in-install)
- adapt patches:
o rediff: patch31 (syslog), patch64 (vendor popt), patch84 (rpmqv-ghost),
patch116 (qv-use-same-indentation), patch49, patch56, patch82, patch86,
patch112, patch119
o patch3 (prefer-pic in "file" lib .la)
o patch17 (keep enhancements from upstream version)
o patch114 (read our macros files)
o patch124 (speedup-by-not-checking-same-files-with-different-paths-through-symlink)
and add patch1124 (revert-unused-skipDir-functionality) needed by patch124
- use autoreconf and don't do it in db/dist (it fails)
- drop perl-RPM (unused and no more in rpm 4.4.2.2)
- drop now unneeded perldirs patch85
- don't build apidocs since we don't bundle them
- drop patch78 (Do not use futex, but fcntl) which is not doing anything anymore
- fix rpm allowing to remove a package which is both suggested and required (#34342)
- remove global RPMLOCK, it doesn't seem to fix rpmdb issues, and may cause
installer to crash
- patch global-RPMLOCK:
o fix rebuilddb (#34181)
o don't lock when querying the db as non-superuser
o fix locking in shared mode (ie fix typo)
o remove message "RPMLOCK_NOWAIT should have been set!"
which can't be done reliably
o more debug messages about RPMLOCK
- patch global-RPMLOCK:
o allow rpm to not wait on lock when RPMLOCK_NOWAIT is set
o ensure we don't have a dead-lock in package scriptlets which call "rpm -q"
(eg: dkms). but it will die with an error
o do create /var/lib/rpm directory when needed (as done in rpmdb/db3.c)
o use flock instead of fcntl+F_SETLK
(so that we really ensure the same process doesn't open twice the db)
(still need not to lock when accessing in shared mode with not enough rights
to write RPMLOCK)
- fix creating /RPMLOCK instead of /var/lib/rpm/RPMLOCK
- do not use __db* files anymore, make them private
- (re-)add a global lock to ensure everything is correct
(to ensure doing "rpm -qa" is always correct even if an upgrade is in
progress)
- fix rpm -K segfaulting on corrupted header (#33735)
- make "rpm -bb --quiet" quiet as should be
- bug fix release, fix stupid typo in /etc/platform32 reading patch
- really allow to use "linux32 rpm -bb" instead of "linux32 rpm -bb --target x86_64"
(it uses /etc/rpm/platform32)
- allow to use "linux32 rpm -bb" instead of "linux32 rpm -bb --target x86_64"
(for this to work, %{_host_cpu32} must be set in rpm-mandriva-setup)
- fix popt.h (remove N_) (#31397)
- add patch to speedup simple "rpm -e" or "rpm -U"
- make find-lang --with-gnome picks up omf files (rhbz#251400) (rpm.org patch)
- add the bug number fixed
- patch fixing parse_hdlist (and so genhdlist2) on heavy loaded boxes
- fix segfault on weird rpm that used to work (#32102)
- remove recursive %_mandir (and don't keep it, it's not needed anymore)
- do not use nptl on any arch (since x86_64 specific issues seem to disappear
when using same locking mechanism as x86) (may fix #32102)
- fix segfault occuring with some old rpm v3 from urpmi test cases
- allow building --with debug
- restore do-check-free-size-when-bavail-is-0 patch (after rediff)
- make it explicit that older perl-URPM are not compatible with 4.4.8
- add patches 112 113 114 115 from both rpm 4.4.9 and hg.rpm.org
- cleanup provides-obsoleted.patch, moving other fixes into rebuilddb-with-root.patch
- more complete fix-segfault-on-header-without-RPMTAG_NAME patch (from upstream)
- don't segfault on header without RPMTAG_NAME
- add requires pkgconfig in rpm-build (#30632)
(it is needed for performing automatic computation of pkgconfig files
dependencies, see /usr/lib/rpm/mandriva/pkgconfigdeps.sh)
- add patch (ensure-uninst-callback-is-called-for-empty-packages)
to allow urpmi to display "remove package foo" when removing an empty package
+ Frederic Crozat
- bump popt release too
- Bump minimal version of rpm-mandriva-setup, to ensure filetriggers
are enabled during upgrade from older distributions
+ Oden Eriksson
- bump release for popt as well
- fix deps and rebuild against latest neon-devel
- don't provide these files twice:
/usr/lib/rpm/rpmdb_deadlock
/usr/lib/rpm/rpmdb_dump
/usr/lib/rpm/rpmdb_load
/usr/lib/rpm/rpmdb_loadcvt
/usr/lib/rpm/rpmdb_stat
/usr/lib/rpm/rpmdb_svc
/usr/lib/rpm/rpmdb_verify
they are now only provided in the main rpm package, and not
also in the devel package. that is the way to go according to
jbj.
+ Toshihiro Yamagishi
- get rid of _host_vendor definition. It should be defined by rpm-xxx-macros.
- use original find-requires,find-provides,find-lang when build with turbolinux.
- headerIconv.patch: check realloc() value
- convert the data to a specific encoding which used in the selected locale.
+ Olivier Thauvin
- fix #31535: lowercase OS tag to compare against platform
- use upstream patch
- fix config file replace when not yet in db
- kill not bad macros definition
- find-lang is now a perl script
- fix buildreq for autoconf/automake
- apply new devel policy
- kill coloring patch, buggy, and not so usefull
- patch116: fix rpm -V output (#31287)
- patch115: avoid breakage of conectiva --without bm feature
- remove patch 110, add patch114: do no longer use rpmrc, using only macros instead, so requiring latest rpm-mandriva-setup to have /etc/rpm/platform
- fix perl module release
- patch to backport _docdir_fmt macro
- 4.4.8
o rediff patch 72, 87, 49, 62, 17, 56, 64
o kill patch (merge or fix upstream, woot):
57, 80, 90, 93, 193, 95, 96, 97, 99, 103, 104, 105, 109, 110, 111,
112, 114, 115, 101, 102, 107
o kill patch 69, hard to maintain, will be merge upstream
o Patch109: workaround specfile parsing check,
http://rpm5.org/cvs/tktview?tn=6
o Patch110: still read rpmrc (this should die one day)
o Patch111: trim twice / in buildroot, making unpackaged files check failing
o Patch112: avoid a very issue regarding Fopen()
+ Adam Williamson
- rebuild against new beecrypt