Here is a patch which adds a --orphans (-o) option to qfile. It inverses the purpose of the command: instead of listing packages owning the files, it lists files which don't have any owner. Exit code is zero if no orphan found. An example of use is to look for orphan libtool files (left behind because of fix_libtool_files.sh): % find /lib /usr/lib -type f -name "*.la" -print0 | xargs -0 -n 500 qfile -o This patch is based on the optimized qfile from bug #139825, it should be applied after attachment #91385 [details, diff] (on portage-utils-0.1.19). This way, the above command takes ~8s (for ~1500 .la files).
Created attachment 91495 [details, diff] qfile-0.1.19-orphans.patch
Please assign portage-utils related bugs to portage-utils@ vs tools-portage@
(In reply to comment #2) > Please assign portage-utils related bugs to portage-utils@ vs tools-portage@ > If only i could... The "Assign To" field is no more editable by users since a few months. It's now only filled automatically, depending on the component you select. Hence the CC i added for you to pick it. But yes, next time i'll leave it without a component, so that it's wranglers who get the spam and not portage-tools@.
Patch failed to apply cleanly to CVS revision 1.31 of qfile.c I fixed up the rejects and committed it as cvs revision: 1.32
I want to revert this patch as it causes segfaults. $ ./q file /bin/* > /dev/null Segmentation fault $ cvs diff -r 1.31 qfile.c > qfile-1.32.diff $ cat qfile-1.32.diff | patch -R -p0 $ ./q file /bin/* > /dev/null $ echo $? 0 Added patch back and strace shows. 1576 open("/var/db/pkg/app-arch/bzip2-1.0.3-r5/CONTENTS", O_RDONLY) = 5 1576 fstat64(5, {st_mode=S_IFREG|0644, st_size=2024, ...}) = 0 1576 mmap2(NULL, 131072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x527c2000 1576 read(5, "dir /bin\nsym /bin/bzegrep -> bzg"..., 131072) = 2024 1576 fstat64(1, {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 3), ...}) = 0 1576 ioctl(1, SNDCTL_TMR_TIMEBASE or TCGETS, 0x5c34d780) = -1 ENOTTY (Inappropriate ioctl for device) 1576 mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x527c1000 1576 --- SIGSEGV (Segmentation fault) @ 0 (0) --- 1576 +++ killed by SIGSEGV +++ We will have to solve this before the release of .20
Here is a little debug info. solar@onyx /var/cvsroot/gentoo-projects/portage-utils $ gdb -q ./q Using host libthread_db library "/lib/libthread_db.so.1". gdb> run file /bin/* profile ../usr/portage/profiles/hardened/x86/2.6/make.defaults profile /etc/make.globals profile /etc/make.conf ACCEPT_LICENSE = * INSTALL_MASK = ARCH = CONFIG_PROTECT = /usr/lib/X11/xkb NOCOLOR = 0 FEATURES = noman noinfo nodoc sandbox distlocks metadata-transfer buildpkg noinfo nodoc -splitdebug -installsources confcache -distclean cvs PORTDIR = /usr/portage PORTAGE_BINHOST = ftp://tinderbox.x86.dev.gentoo.org/hardened/x86/ PORTAGE_TMPDIR = /var/tmp PKGDIR = /usr/portage/packages/x86/ ROOT = / q: lookup_applet(): found applet q at 0x804efd0 q: lookup_applet(): Looking up applet (file) by short name q: lookup_applet(): found applet by short name qfile file: qfile_main(): argc=145 argv[0]=file argv[1]=/bin/Mail app-arch/bzip2 (/bin/bzegrep) Program received signal SIGSEGV, Segmentation fault. _______________________________________________________________________________ eax:00000000 ebx:080C1DD0 ecx:080C1E38 edx:000000A1 eflags:00010202 esi:0000000B edi:BFFFA5A9 esp:BFFFA4D0 ebp:BFFFB9C8 eip:08051BF4 cs:0073 ds:007B es:007B fs:0000 gs:0000 ss:007B o d I t s z a p c [007B:BFFFA4D0]---------------------------------------------------------[stack] BFFFA500 : 18 04 0A 08 78 1D 0A 08 - 44 28 07 08 01 00 00 00 ....x...D(...... BFFFA4F0 : C3 1E 0A 08 00 00 00 00 - A8 08 0A 08 60 06 0A 08 ............`... BFFFA4E0 : 2A 1E 0C 08 3B 2A 06 08 - 00 00 00 00 00 00 00 00 *...;*.......... BFFFA4D0 : D0 1D 0C 08 A4 A5 FF BF - 21 1E 0C 08 40 E6 06 08 ........!...@... [007B:BFFFA5A9]---------------------------------------------------------[ data] BFFFA5A9 : 62 7A 65 67 72 65 70 00 - 2D 3E 20 62 7A 67 72 65 bzegrep.-> bzgre BFFFA5B9 : 70 00 31 31 33 32 33 33 - 32 30 31 38 00 00 66 66 p.1132332018..ff [0073:08051BF4]---------------------------------------------------------[ code] 0x8051bf4 <qfile+596>: movw $0x1,(%eax,%esi,2) 0x8051bfa <qfile+602>: inc %esi 0x8051bfb <qfile+603>: jmp 0x8051b60 <qfile+448> 0x8051c00 <qfile+608>: mov 0xffffeb40(%ebp),%edx 0x8051c06 <qfile+614>: mov 0x8(%edx),%eax 0x8051c09 <qfile+617>: mov %eax,(%esp) ------------------------------------------------------------------------------ 0x08051bf4 in qfile (path=0x80a1d78 "/var/db/pkg/app-arch", argc=0x90, base_names=0x80a0418, dir_names=0x80a0660, real_dir_names=0x80a08a8, non_orphans=0x0) at qfile.c:130 130 non_orphans[i] = 1;
And a backtrace. gdb> bt full #0 0x08051bf4 in qfile (path=0x80a1d78 "/var/db/pkg/app-arch", argc=0x90, base_names=0x80a0418, dir_names=0x80a0660, real_dir_names=0x80a08a8, non_orphans=0x0) at qfile.c:130 e = (contents_entry *) 0x8072844 fp = (FILE *) 0x80c1e70 dir = (DIR *) 0x80a1d98 dentry = (struct dirent *) 0x0 entry_basename = 0xbfffa5a9 "bzegrep" entry_dirname = 0x80c1e00 "" p = 0x80c1e04 "" buf = "sym /bin/bzegrep\000-> bzgrep\0001132332018\000\000ffd0957d78a87d437a498a24b5\0001147185899\000\0002e4ff354b800\0001147185899\000\000\000\000\000\000\000\000\0009\000\000\000\000\0001671137\000\00020\000\000232\000\000\000\000\0009\000\0007171\000\000\000667171\000\0001\000\0001\000\000\000\00071\000\000@:
And a backtrace. gdb> bt full #0 0x08051bf4 in qfile (path=0x80a1d78 "/var/db/pkg/app-arch", argc=0x90, base_names=0x80a0418, dir_names=0x80a0660, real_dir_names=0x80a08a8, non_orphans=0x0) at qfile.c:130 e = (contents_entry *) 0x8072844 fp = (FILE *) 0x80c1e70 dir = (DIR *) 0x80a1d98 dentry = (struct dirent *) 0x0 entry_basename = 0xbfffa5a9 "bzegrep" entry_dirname = 0x80c1e00 "" p = 0x80c1e04 "" buf = "sym /bin/bzegrep\000-> bzgrep\0001132332018\000\000ffd0957d78a87d437a498a24b5\0001147185899\000\0002e4ff354b800\0001147185899\000\000\000\000\000\000\000\000\0009\000\000\000\000\0001671137\000\00020\000\000232\000\000\000\000\0009\000\0007171\000\000\000667171\000\0001\000\0001\000\000\000\00071\000\000@:¾\a@¸\n\002\000H\005\b\bà\003\b\b Y\023@Ì=\023@à\003\b\b Y\023@"... pkg = "app-arch/bzip2-1.0.3-r5\000-r11\0004.2\000\0001\0001\0001\000 ¥ÿ¿t¥ÿ¿±ª\a@ ¥ÿ¿uD¢Ã'æ\006\b\000\000\000\000'æ\006\bÌ=\023@\020\000\000\000\030æ\006\b\234¦ÿ¿}R\a@ ¥ÿ¿n1\006\bЦÿ¿\030æ\006\b\000\000\000\000@¦ÿ¿Ì¦ÿ¿°\006\a\b ¥" atom = (depend_atom *) 0x80c1dd0 i = 0xb path_ok = 0x1 #1 0x080526de in qfile_main (argc=0x91, argv=0xbfffeb48) at qfile.c:247 tmppath = "/bin", '\0' <repeats 4091 times> abspath = "/bin\000znew", '\0' <repeats 4086 times> dir = (DIR *) 0x80803e0 dentry = (struct dirent *) 0x0 i = 0x90 p = 0x80a1d78 "/var/db/pkg/app-arch" basenames = (char **) 0x80a0418 dirnames = (char **) 0x80a0660 realdirnames = (char **) 0x80a08a8 pwd = 0xbffffca4 "/var/cvsroot/gentoo-projects/portage-utils" non_orphans = (short int *) 0x0 search_orphans = 0x8a8 __FUNCTION__ = "qfile_main" #2 0x0804f333 in q_main (argc=0x92, argv=0xbfffeb48) at q.c:132 buf = '\0' <repeats 1392 times>, "hàÿ¿\000`\023@´o\001@(âÿ¿o[\000@\000`\023@Ô\034\000\000\003\000\000\0002\000\000\000ÿÿÿÿ", '\0' <repeats 45 times>, " \021\000À\224\021\000À\224\021\000\000\000\000\000\005\000\000\000\000 \021\000\000à\021\000\020Ò\021\000Ôü\021\000\000\220\021\000\003\000\000\000\216\002\001@", '\0' <repeats 12 times>, "\216\002\001@\000\000\000\000\000\000\000\000´o\001@\000\000\000\000\000\000\000\000¼àÿ¿´o\001@\017\000\000\000\020z\001@Ìàÿ¿å\003\001@\b\000\000\000\017\000\000\000´o\001@\fáÿ¿µ\215\000@h|\001@\000z\001@\017\000\000\000h|\001"... i = 0x0 p = 0x1 <Address 0x1 out of bounds> func = (APPLET) 0x80c1e38 __FUNCTION__ = "q_main" #3 0x0804f43b in main (argc=0x0, argv=0xbfffeb44) at main.c:1011 No locals. gdb>
$ ./qfile /bin/bzegrep Segmentation fault $ ls -l /bin/bzegrep lrwxrwxrwx 1 root root 6 Nov 18 2005 /bin/bzegrep -> bzgrep $ ltrace -o woof ./qfile /bin/bzegrep $ tail -n50 woof __strtol_internal("1147185899", NULL, 10) = 1147185899 strrchr("/usr/share/man/man1/upx.1.gz", '/') = "/upx.1.gz" strcmp("upx.1.gz", "bzegrep") = 1 fgets("obj /usr/share/man/man1/upx.1.gz", 1024, 0x80a3cc8) = NULL fclose(0x80a3cc8) = 0 readdir(0xb7fdf008) = 0xb7fdf128 asprintf(0xbfffa52c, 0x80637bd, 0x80a3c40, 0xb7fdf133, 0xb7fdf11b) = 44 fopen("/var/db/pkg/app-arch/bzip2-1.0.3"..., "r") = 0x80a3cc8 free(0x80a3bd8) = <void> __xpg_basename(0x80a3c40, 0x8062715, 0x80a3c40, 0xb7fdf133, 0xb7fdf11b) = 0x80a3c4c snprintf("app-arch/bzip2-1.0.3-r5", 126, "%s/%s", "app-arch", "bzip2-1.0.3-r5") = 23 fgets("dir /bin\n", 1024, 0x80a3cc8) = 0xbfffa5b0 strrchr("dir /bin\n", '\n') = "\n" strrchr("/bin", '/') = "/bin" strcmp("bin", "bzegrep") = -1 fgets("sym /bin/bzegrep -> bzgrep 11323"..., 1024, 0x80a3cc8) = 0xbfffa5b0 strrchr("sym /bin/bzegrep -> bzgrep 11323"..., '\n') = "\n" strrchr("/bin/bzegrep -> bzgrep 113233201"..., ' ') = " 1132332018" strstr("/bin/bzegrep -> bzgrep", " -> ") = " -> bzgrep" __strtol_internal("1132332018", NULL, 10) = 1132332018 strrchr("/bin/bzegrep", '/') = "/bzegrep" strcmp("bzegrep", "bzegrep") = 0 __strdup(0xbfffa5b4, 0, 10, 0, 0xeac22f0a) = 0x80a3e38 strrchr("/bin/bzegrep", '/') = "/bzegrep" strcmp("/bin", "/bin") = 0 free(0x80a3e38) = <void> strlen("app-arch/bzip2-1.0.3-r5") = 23 malloc(105) = 0x80a3e50 memcpy(0x80a3ea1, "app-arch/bzip2-1.0.3-r5", 23) = 0x80a3ea1 strrchr("app-arch/bzip2-1.0.3-r5", '/') = "/bzip2-1.0.3-r5" strrchr("app-arch", '/') = NULL strstr("bzip2-1.0.3-r5", ".ebuild") = NULL strlen("bzip2-1.0.3-r5") = 14 __ctype_b_loc() = 0x40137938 __strtol_internal("5", NULL, 10) = 5 strchr("bzip2-1.0.3", '_') = NULL strlen("bzip2-1.0.3") = 11 __ctype_b_loc() = 0x40137938 __ctype_b_loc() = 0x40137938 __ctype_b_loc() = 0x40137938 __ctype_b_loc() = 0x40137938 __ctype_b_loc() = 0x40137938 strlen("1.0.3") = 5 sprintf("1.0.3-r5", "%s-r%i", "1.0.3", 5) = 8 sprintf("bzip2-1.0.3-r5", "%s-%s", "bzip2", "1.0.3-r5") = 14 printf("%s%s/%s%s%s", "\033[00;01m", "app-arch", "\033[36;01m", "bzip2", "\033[00;00m") = 38 printf(" (%s)\n", "/bin/bzegrep") = 16 free(0x80a3e50) = <void> --- SIGSEGV (Segmentation fault) --- +++ killed by SIGSEGV +++
Created attachment 92150 [details, diff] qfile-orphans-fix.patch Oops, sorry for the breakage... and thanks for your debug infos. Here is a patch which fixes it (the 2nd chunk). It also fixes a bug on "/foo" queries (1st chunk), and change output of --orphans (3rd chunk), to report exactly whatever the query arguments was (otherwise, queries like "./foo" were looking weird, since they were reported as "${PWD}/./foo"). The patch is for CVS 1.32.
cvs ci -m '- fix up segv in initial orphan handling code. attachment#92150 [details, diff] from bug 140019 degrenier@easyconnect' qfile.c /var/cvsroot/gentoo-projects/portage-utils/qfile.c,v <-- qfile.c new revision: 1.33; previous revision: 1.32
Released in portage-utils-0.1.20
TGL, I'd like to say thank you for this option. It's starting to prove very useful to me. I'm think I can use it to test for binary transactions before installing a binary package perhaps doing something like this. qfile -o $(scanelf -n -lq -F%n#f| tr , '\n' | sort -u)
(In reply to comment #12) > I'm think I can use it to test for binary transactions before > installing a binary package FYI, i think it won't really work without some false positives (i mean, without even talking of weird cases where a lib is package.provided and manually installed, or things like that) Actually, I've just realized that while writing a script which was trying to do revdep-rebuild's job faster, by just looking at the NEEDED files in the vdb, and then calling "qfile -o" on the needed libs to detect the missing ones. It happened to break on cases like this one, for which i see no good workaround: - "net-libs/liblockfile" installs "liblockfile.so.1.0" and a "liblockfile.so" symlink, but no "liblockfile.so.1", - ldconfig (i guess), when ran by portage, adds the missing "liblockfile.so.1" symlink, - "mail-client/sylpheed-claws" links against "liblockfile.so.1", - qfile can't find "liblockfile.so.1" in any CONTENTS files, and thus the script reports "sylpheed-claws" as a candidate for the rebuild. Or maybe there could be a QA rule stating that ebuilds should always provide all the symlinks for the libs they install, but it sounds hard to justify since nothing is actually broken in the above "liblockfile" example.
I think procmail does it also. I've scanned hundreds of packages using this method and fixed a few of the ones I've found that were broken. And yeah you are quite right it should become a QA problem at that point. Side note this is what I've been using. http://dev.gentoo.org/~solar/portage_misc/bashrc.verify-needed
*** Bug 27652 has been marked as a duplicate of this bug. ***
