- Pacman-dev - lists.archlinux.org

[pacman-dev] New way to download signing keys prone to MITM attacks?
by Manuel Reimer 09 Feb '15

09 Feb '15

Hello, today, I was asked for the first time whether I want to download a signing key. So far this was done using a "keyring" package, which, itself, was signed using a trusted key. How do you prevent MITM attacks? For me this seems like anyone, who can perform a MITM attack, can trick me into installing virtually any package as long as he signs it with a key somewhere available on a public keyserver. Of course I would be asked whether I want to import that key but how do I know if the key is really valid and trusted? My guess is that most users will just say "yes" in this case. For me this seems to be a big step backwards in terms of security. Please correct me if I'm wrong. Thanks in advance. Manuel

2 4

[pacman-dev] [PATCH] Add AM_MAINTAINER_MODE to configure.ac and disable by default
by Allan McRae 09 Feb '15

09 Feb '15

We had issues with people trying to build the 4.2.0 tarball while not having automake-1.14 installed. This appears to be due to timestamp issues resulting in autotools trying to regenerate various files and then the Makefiles deciding they need specific versions of automake. This can be fixed by adding AM_MAINTAINER_MODE([disable]) to configure.ac, which means that any build will not try and rebuild any build scripts unless explicitly requested. Pacman developers can consider using --enable-maintainer-mode when running configure to use autotools to update build scripts as needed. Signed-off-by: Allan McRae <allan(a)archlinux.org> --- That is a long explanation for a single line change! configure.ac | 1 + 1 file changed, 1 insertion(+) diff --git a/configure.ac b/configure.ac index e0ed3cf..f17ae89 100644 --- a/configure.ac +++ b/configure.ac @@ -62,6 +62,7 @@ AC_REQUIRE_AUX_FILE([tap-driver.sh]) AC_CANONICAL_HOST AM_INIT_AUTOMAKE([1.11 foreign]) AM_SILENT_RULES([yes]) +AM_MAINTAINER_MODE([disable]) LT_INIT LIB_VERSION=`expr lib_current - lib_age`.lib_age.lib_revision -- 2.2.2

3 5

[pacman-dev] [PATCH] paccache: added --quiet option
by Maxim Andersson 09 Feb '15

09 Feb '15

Less output, although errors and warnings will still be printed (scripts/library/output_format.sh). Cleaner to have -q,--quiet than >/dev/null in cron. Signed-off-by: Maxim Andersson <thesilentboatman(a)gmail.com> --- contrib/paccache.sh.in | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/contrib/paccache.sh.in b/contrib/paccache.sh.in index 1690583..02fae52 100644 --- a/contrib/paccache.sh.in +++ b/contrib/paccache.sh.in @@ -27,6 +27,7 @@ declare -a cachedirs=() candidates=() cmdopts=() whitelist=() blacklist=() declare -i delete=0 dryrun=0 filecount=0 move=0 needsroot=0 totalsaved=0 verbose=0 declare delim=$'\n' keep=3 movedir= scanarch= +QUIET=0 USE_COLOR='y' m4_include(../scripts/library/output_format.sh) @@ -135,6 +136,8 @@ runcmd() { } summarize() { + (( QUIET )) && return + local -i filecount=$1; shift local seenarch= seen= arch= name= local -r pkg_re='(.+)-[^-]+-[0-9]+-([^.]+)\.pkg.*' @@ -192,6 +195,7 @@ Usage: ${myname} <operation> [options] [targets...] delimited. -k, --keep <num> keep "num" of each package in the cache (default: 3). --nocolor remove color from output. + -q, --quiet minimize output -u, --uninstalled target uninstalled packages. -v, --verbose increase verbosity. specify up to 3 times. -z, --null use null delimiters for candidate names (only with -v @@ -205,9 +209,9 @@ version() { echo 'Copyright (C) 2011 Dave Reisner <dreisner(a)archlinux.org>' } -OPT_SHORT=':a:c:dfhi:k:m:rsuVvz' +OPT_SHORT=':a:c:dfhi:k:m:qrsuVvz' OPT_LONG=('arch:' 'cachedir:' 'dryrun' 'force' 'help' 'ignore:' 'keep:' 'move' - 'nocolor' 'remove' 'uninstalled' 'version' 'verbose' 'null') + 'nocolor' 'quiet' 'remove' 'uninstalled' 'version' 'verbose' 'null') if ! parseopts "$OPT_SHORT" "${OPT_LONG[@]}" -- "$@"; then exit 1 @@ -247,11 +251,13 @@ while :; do keep=$(( 10#$keep )) fi shift ;; - --nocolor) - USE_COLOR='n' ;; -m|--move) move=1 movedir=$2 shift ;; + --nocolor) + USE_COLOR='n' ;; + -q|--quiet) + QUIET=1 ;; -r|--remove) delete=1 ;; -u|--uninstalled) -- 2.3.0

3 3

[pacman-dev] [libalpm] Installing and removing a package in the same handle lifecycle
by Fabien Bourigault 08 Feb '15

08 Feb '15

Hi, I'm a PackageKit alpm backend contributor and I found a strange behavior when removing a package in the same alpm_handle_t lifecycle. Here you have the simplified workflow : alpm initialization new transaction install inkscape end transaction new transaction remove inkscape end transaction When removing the package, backend does an alpm_db_get_pkg (localdb, name) to get the package and then call alpm_remove_pkg with this found package. alpm_remove_pkg returns an error with 6 as errno. I looked in alpm_remove_pkg code and there is 3 cases returning such error. The one failing is pkg coming from localdb. I have no knowledge of libalpm internals but from the outside it looks like the package structure was copied ot the localdb in installation phase but the db field was not updated. I hope this is a bug that could be solved. Regards, Fabien

2 3

[pacman-dev] [PATCH] zsh completion: get installed packages with pacman -Qq
by Roma Kovalev 08 Feb '15

08 Feb '15

Currently _pacman_completions_installed_packages() globs the directory @localstatedir@/lib/pacman/local/, this does not respect changes to DBPath in /etc/pacman.conf, which cause completion to stop working. Instead of globbing $(pacman -Qq) can be called to get a list of installed packages correctly. Signed-off-by: Roma Kovalev <roma.kvl(a)gmail.com> --- contrib/zsh_completion.in | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/contrib/zsh_completion.in b/contrib/zsh_completion.in index 1cae1de..0eae5e9 100644 --- a/contrib/zsh_completion.in +++ b/contrib/zsh_completion.in @@ -285,9 +285,8 @@ _pacman_completions_installed_groups() { # provides completions for installed packages _pacman_completions_installed_packages() { - local -a cmd packages packages_long - packages_long=(@localstatedir@/lib/pacman/local/*(/)) - packages=( ${${packages_long#@localstatedir@/lib/pacman/local/}%-*-*} ) + local -a packages + packages=( $(pacman -Qq) ) compadd "$@" -a packages } -- 2.2.2

3 2

[pacman-dev] [PATCH] Disable directory ownership warnings
by Allan McRae 04 Feb '15

04 Feb '15

There was a lot of confusion regarding these warnings, particularly for packages that create users post_install and then chown the directories. Signed-off-by: Allan McRae <allan(a)archlinux.org> --- lib/libalpm/add.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/lib/libalpm/add.c b/lib/libalpm/add.c index 82461ee..a6b83f5 100644 --- a/lib/libalpm/add.c +++ b/lib/libalpm/add.c @@ -231,6 +231,12 @@ static int extract_single_file(alpm_handle_t *handle, struct archive *archive, gid_t entrygid = archive_entry_gid(entry); /* case 6: existing dir, ignore it */ + +#if 0 + /* Disable these warnings until our user management has improved. Currently + many packages have to create users in post_install and chown the directories. + These all resulted in "false-positive" warnings. */ + if(lsbuf.st_mode != entrymode) { /* if filesystem perms are different than pkg perms, warn user */ mode_t mask = 07777; @@ -252,6 +258,7 @@ static int extract_single_file(alpm_handle_t *handle, struct archive *archive, "filesystem: %u:%u package: %u:%u\n", filename, lsbuf.st_uid, lsbuf.st_gid, entryuid, entrygid); } +#endif _alpm_log(handle, ALPM_LOG_DEBUG, "extract: skipping dir extraction of %s\n", filename); -- 2.2.2

2 2

[pacman-dev] Upgrading packages on case-insensitive filesystem
by David Macek 04 Feb '15

04 Feb '15

Hi all. The issue is that pacman refuses to replace a file during package upgrade if the case of the file's name has changed between versions. The error displayed is this: > error: failed to commit transaction (conflicting files) > example2: /path/to/file exists in filesystem > Errors occurred, no packages were upgraded. I'm using pacman on Windows/NTFS (ported as part of the MSYS2 project). I'm not completely sure if the problem is in upstream pacman, or only in the ported one, but it's happened enough times to prompt me to ask you people. I don't fully understand _alpm_db_find_fileconflicts, but it seems that it can't discover that the conflicting file will be removed with the old version of the package before an actual conflict can occur. Would this be considered as a bug in pacman? For the whole log (not much longer) and example PKGBUILDs, look here: https://gist.github.com/elieux/8faaf423601d7ef01d0f Note that NTFS while is not case-sensitive by default, it is case-preserving. -- David Macek

3 2

[pacman-dev] updpkgsums deletes more variables than it should
by David Macek 03 Feb '15

03 Feb '15

Hi all. It seems to me that updpkgsums destroys variables that it should not, specifically those with a suffix, e.g. "md5sums_extra". This is the regex in question: /^[[:blank:]]*(md|sha)[[:digit:]]+sums(_[^=]+)?=/,/\)[[:blank:]]*(#.*)?$/ I couldn't find anything that would suggest that such variable names are reserved for pacman. -- David Macek

4 4

[pacman-dev] [RFC] Correct directory permissions if the extracting package is the only owner
by Allan McRae 03 Feb '15

03 Feb '15

Adjusting permissions of directories in their install scripts results in warnings such as the following when the package update: warning: directory ownership differs on /var/lib/postfix/ filesystem: 73:0 package: 0:0 If the package being installed is the only owner of the directory, update the directory's permissions to those in the package file. This also allows directory permissions for a package to be changed directly in the PKGBUILD file rather that requiring adjustment in an install scriptlet. Signed-off-by: Allan McRae <allan(a)archlinux.org> --- This does require searching through all installed packages filelist to see if any other directories own this file, which does create a slowdown in these error cases. Once the ability to use symbolic names for owners is added, this code path will run only if there is an acutal error or in the rare case that directory permissions are genuinely being changed. Packages can currently avoid this being run by reverting their directory permission changes in pre_upgrade(). I'd like to push this to maint if people agree. These "false" warnings are causing too many issues. lib/libalpm/add.c | 61 +++++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 46 insertions(+), 15 deletions(-) diff --git a/lib/libalpm/add.c b/lib/libalpm/add.c index 82461ee..60db22f 100644 --- a/lib/libalpm/add.c +++ b/lib/libalpm/add.c @@ -167,6 +167,20 @@ static int extract_db_file(alpm_handle_t *handle, struct archive *archive, return perform_extraction(handle, archive, entry, filename); } +static int other_directory_owners(alpm_handle_t *handle, const char *entryname, const char *pkgname) { + alpm_list_t *packages = alpm_db_get_pkgcache(handle->db_local); + alpm_list_t *i; + + for(i = packages; i ; i = alpm_list_next(i)) { + if(alpm_filelist_contains(alpm_pkg_get_files(i->data), entryname) && + strcmp(((alpm_pkg_t *)(i->data))->name, pkgname) != 0) { + return 1; + } + } + + return 0; +} + static int extract_single_file(alpm_handle_t *handle, struct archive *archive, struct archive_entry *entry, alpm_pkg_t *newpkg, alpm_pkg_t *oldpkg) { @@ -232,25 +246,42 @@ static int extract_single_file(alpm_handle_t *handle, struct archive *archive, /* case 6: existing dir, ignore it */ if(lsbuf.st_mode != entrymode) { - /* if filesystem perms are different than pkg perms, warn user */ mode_t mask = 07777; - _alpm_log(handle, ALPM_LOG_WARNING, _("directory permissions differ on %s\n" - "filesystem: %o package: %o\n"), filename, lsbuf.st_mode & mask, - entrymode & mask); - alpm_logaction(handle, ALPM_CALLER_PREFIX, - "warning: directory permissions differ on %s\n" - "filesystem: %o package: %o\n", filename, lsbuf.st_mode & mask, - entrymode & mask); + if(other_directory_owners(handle, entryname, newpkg->name) == 0) { + /* no other package owns this directory, "extract" with correct permissions */ + if(chmod(filename, entrymode) != 0) { + goto warn_permissions; + } + } else { +warn_permissions: + /* if filesystem perms are different than pkg perms, warn user */ + _alpm_log(handle, ALPM_LOG_WARNING, _("directory permissions differ on %s\n" + "filesystem: %o package: %o\n"), filename, lsbuf.st_mode & mask, + entrymode & mask); + alpm_logaction(handle, ALPM_CALLER_PREFIX, + "warning: directory permissions differ on %s\n" + "filesystem: %o package: %o\n", filename, lsbuf.st_mode & mask, + entrymode & mask); + + } } if((entryuid != lsbuf.st_uid) || (entrygid != lsbuf.st_gid)) { - _alpm_log(handle, ALPM_LOG_WARNING, _("directory ownership differs on %s\n" - "filesystem: %u:%u package: %u:%u\n"), filename, - lsbuf.st_uid, lsbuf.st_gid, entryuid, entrygid); - alpm_logaction(handle, ALPM_CALLER_PREFIX, - "warning: directory ownership differs on %s\n" - "filesystem: %u:%u package: %u:%u\n", filename, - lsbuf.st_uid, lsbuf.st_gid, entryuid, entrygid); + if(other_directory_owners(handle, entryname, newpkg->name) == 0) { + /* no other package owns this directory, "extract" with correct owners */ + if(chown(filename, entryuid, entrygid) != 0) { + goto warn_owners; + } + } else { +warn_owners: + _alpm_log(handle, ALPM_LOG_WARNING, _("directory ownership differs on %s\n" + "filesystem: %u:%u package: %u:%u\n"), filename, + lsbuf.st_uid, lsbuf.st_gid, entryuid, entrygid); + alpm_logaction(handle, ALPM_CALLER_PREFIX, + "warning: directory ownership differs on %s\n" + "filesystem: %u:%u package: %u:%u\n", filename, + lsbuf.st_uid, lsbuf.st_gid, entryuid, entrygid); + } } _alpm_log(handle, ALPM_LOG_DEBUG, "extract: skipping dir extraction of %s\n", -- 2.2.2

3 4

[pacman-dev] Release 4.2.1?
by Allan McRae 02 Feb '15

02 Feb '15

Hi all, A number of fixes have built up in maint. What do we need done for a 4.2.1 release? - I'd like comments on my directory permissions patch. - the release tarball embeds automate and aclocal versions so can only be built with the current version of automake. Not sure how to stop this... Anything else? A

1 0