i had read about an incident where someone had used portsclean(1) and lost the base systems gcc-libs, so i used "portsclean -Li" to clean out old libs after an upgrade. the output was: # portsclean -Li ** /usr/local/lib/gcc-4.2.2/libgcc_s.so.1 is shadowed by /lib/libgcc_s.so.1 /lib/libgcc_s.so.1 <- ? /usr/local/lib/gcc-4.2.2/libgcc_s.so.1 <- gcc-4.2.2_20070927 --> This may be an undesirable situation Delete /lib/libgcc_s.so.1 OK? [no] ** /usr/local/lib/gcc-4.2.2/libstdc++.so.6 is shadowed by /usr/lib/libstdc++.so.6 /usr/lib/libstdc++.so.6 <- ? /usr/local/lib/gcc-4.2.2/libstdc++.so.6 <- gcc-4.2.2_20070927 --> This may be an undesirable situation Delete /usr/lib/libstdc++.so.6 OK? [no] ** /usr/local/lib/gcc-4.2.2/libssp.so.0 is shadowed by /lib/libssp.so.0 /lib/libssp.so.0 <- ? /usr/local/lib/gcc-4.2.2/libssp.so.0 <- gcc-4.2.2_20070927 --> This may be an undesirable situation Delete /lib/libssp.so.0 OK? [no] ** /usr/local/lib/gcc-4.2.2/libobjc.so.2 is shadowed by /usr/lib/libobjc.so.2 /usr/lib/libobjc.so.2 <- ? /usr/local/lib/gcc-4.2.2/libobjc.so.2 <- gcc-4.2.2_20070927 --> This may be an undesirable situation Delete /usr/lib/libobjc.so.2 OK? [no] ** /usr/local/lib/gcc-4.2.2/libgomp.so.1 is shadowed by /usr/lib/libgomp.so.1 /usr/lib/libgomp.so.1 <- ? /usr/local/lib/gcc-4.2.2/libgomp.so.1 <- gcc-4.2.2_20070927 --> This may be an undesirable situation Delete /usr/lib/libgomp.so.1 OK? [no] using the default left my system intact: # ll /lib/libgcc_s.so.1 /usr/lib/libstdc++.so.6 /lib/libssp.so.0 /usr/lib/libobjc.so.2 /usr/lib/libgomp.so.1 53 -r--r--r-- 1 root wheel - 56K Sep 24 18:21 /lib/libgcc_s.so.1 55 -r--r--r-- 1 root wheel - 9.5K Sep 24 18:21 /lib/libssp.so.0 1719349 -r--r--r-- 1 root wheel - 29K Sep 24 18:21 /usr/lib/libgomp.so.1 1719378 -r--r--r-- 1 root wheel - 101K Sep 24 18:21 /usr/lib/libobjc.so.2 1719371 -r--r--r-- 1 root wheel - 941K Sep 24 18:21 /usr/lib/libstdc++.so.6 i'm not sure what happens when _not_ using the `-i' switch. the code seems to indicate that nothing gets deleted by default. touching anything in base should be a non-issue IMO, other than adding or changing symlinks for compatibility like perl-ports do. i imaging two solutions: changing portsclean(1) to never touch files in base or change the various gcc-ports to rename their libs in order to rule out these name collisions. the shadowing of vital libs mentioned in portscleans output is indeed undesirable, unless the port compilers hard-code library paths. so my request is to implement either solution, taking care of the shadowing issue as well. Fix: i'd advise to always use "portsclean -Li" to let it query interactively in case libraries with identical names exist in both ports and base. regards, clemens How-To-Repeat: install ports/lang/gcc42 and watch "portsclean -Li".
Responsible Changed From-To: freebsd-ports-bugs->sem Over to maintainer
Sure, it's not pleasant behavior. Passed to my TODO list. -- Dixi. Sem.
State Changed From-To: open->closed Fix committed. Thanks.
sem 2008-01-08 12:33:38 UTC FreeBSD ports repository Modified files: ports-mgmt/portupgrade Makefile Added files: ports-mgmt/portupgrade/files patch-bin-portsclean Log: - A patch from HEAD: when gcc installed, some libraries may be conflicted with system ones. portsclean -Li ask to remove them. No removes by default, but we don't want an user remove system libraries just press 'y' in portscleanup -Li. PR: ports/116830 Reported by: clemens fischer <ino-news at spotteswoode.dnsalias.org> Revision Changes Path 1.244 +1 -0 ports/ports-mgmt/portupgrade/Makefile 1.4 +19 -0 ports/ports-mgmt/portupgrade/files/patch-bin-portsclean (new) _______________________________________________ cvs-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/cvs-all To unsubscribe, send any mail to "cvs-all-unsubscribe@freebsd.org"
sem 2008-01-26 18:29:41 UTC FreeBSD ports repository Modified files: ports-mgmt/portupgrade-devel Makefile distinfo Log: - Update to 2.4.0 * many bug fixes * now all dependencies run under portupgrade control * run 'make config-conditional' and 'make config' with -c and -C options PR: ports/116830, ports/115249 Reported by: clemens fischer <ino-news@spotteswoode.dnsalias.org>, KOMATSU Shinichiro <koma2@lovepeers.org>, and other Revision Changes Path 1.252 +1 -1 ports/ports-mgmt/portupgrade-devel/Makefile 1.176 +3 -3 ports/ports-mgmt/portupgrade-devel/distinfo _______________________________________________ cvs-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/cvs-all To unsubscribe, send any mail to "cvs-all-unsubscribe@freebsd.org"