When calling open(), use O_CLOEXEC as much as possible to ensure the
file descriptor is closed when and if a process using libalpm forks.
For most of these cases, and especially in utility functions, the file
descriptor is opened and closed in the same function, so we don't have
too much to worry about. However, for things like the log file and
database lock file, we should ensure descriptors aren't left hanging
around for children to touch.
This patch is inspired by the problem in FS#36161, where an open file
descriptor to the current working directory prevents chroot() from
working on FreeBSD. We don't need this file descriptor in the child
process, so open it (and now several others) with O_CLOEXEC.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This was a hack done by me in commit d8e88aa017 back in 2007 that is
no longer necessary, given a sufficiently smart compiler and one that
supports the inline keyword.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
filtered_dep was duplicating an alpm_depend_t solely for the purpose of
overriding its depmod and would effectively cause alpm_checkdeps to
ignore ALPM_TRANS_FLAG_NODEPVERSION if the duplication failed. Manually
overriding/restoring the depmod for the original depend removes the
duplication as a point of failure and fixes a memory leak where the
duplicated depend was not being properly freed.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
The gettext functions return a "char *", so do the same for the defines
in the case where gettext is unavailable. This prevents a number of
warnings about const being dropped.
Signed-off-by: Allan McRae <allan@archlinux.org>
Packages can be removed during a sync transaction either directly or
due to conflicts and need to be sorted.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Previously, we only allowed the default of responding to basic auth
challenges. Mirrors requiring authorization are far and away the edge
case, but there's no sense in preventing access to them.
Implements FS#38184.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
We will be adding event structs in the following patches.
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Allowing multiple databases with the same name causes conflicts as they
both point to the same database file but may use different servers,
usages, or siglevels.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
FreeBSD chroot(2) fails with EPERM when a directory file descriptor is open
with either `kern.chroot_allow_open_directories` being 0, or when the process
already is inside a chroot. This is exposed in alpm_run_chroot that uses
opendir() to open a file descriptor to the current directory before doing
the forking and chrooting. Since the file descriptor is not used in the
forked process, we close it.
Fixes FS#36161.
Signed-off-by: Allan McRae <allan@archlinux.org>
This commit:
-- replaces space-based indents with tabs per the coding standards
-- removes extraneous whitespace (e.g. extra spaces between function args)
-- adds missing braces for a one-line if statement
Signed-off-by: Jason St. John <jstjohn@purdue.edu>
This ensures that important events will be logged and consistent
regardless of the frontend. The need for global context in the event
callback is also removed. The event is logged before any post_* scripts
run, so this also moves the post_* script output underneath the event in
the log.
Fixes FS#36504
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
There was a brief window between opening the file descriptor and
creating a stream to it. If the process was interrupted during that
window the lock file would not be removed correctly.
The pid is no longer printed to the lock file as this was virtually
meaningless for lock files on NFS.
Fixes FS#35603
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
This brings file siglevels in line with how db siglevels are handled.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Improves consistency and makes it clear that targs will be modified by
_alpm_recursedeps.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
This allows existing front-ends other than pacman to continue working
normally.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Detecting indirect dependencies by traversing a package's entire
dependency tree is prohibitively slow for larger transactions. Instead
add local packages to the dependency graph. This additionally requires
delaying dependency ordering for sync operations so that removed
packages may be excluded from dependency detection.
tests/sync012.py was also updated to ensure that the dependency cycle
was actually detected.
Fixes FS#37380
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Currently you can manually create and then install a package with a
version not containing a pkgrel. The created local database entry is
invalid as the directory name can not be split by _alpm_splitname due
to the assumtion of hyphens separating name-pkgver-pkgrel.
Ensure the package has a valid version when it is loaded. Fixes FS#35514.
Signed-off-by: Allan McRae <allan@archlinux.org>
On 32bit systems, the progress bar intergrity checking can show values
greater than 100% with large transactions. This is due to the total
size of all package files being greater than a size_t. Use uint64_t
for these sizes.
Fixes FS#36608
Signed-off-by: Allan McRae <allan@archlinux.org>
If any package in a sync transaction is missing a required signature,
we give an uninformative error message (which may or may not state that
the missing signature is the issue). Always output the package with
the missing signature.
Signed-off-by: Allan McRae <allan@archlinux.org>
Function pointer gets uselessly compared for NULL in
every iteration. Move the condition to do it just once.
Signed-off-by: slavomir vlcek <svlc@inventati.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
If the server redirects from ${repo}.db to ${repo}.db.tar.gz pacman gets
this wrong: It saves to new filename and fails when accessing
${repo}.db.
We need the remote filename only when downloading remote files with
pacman's -U operation. This introduces a new field 'trust_remote_name'
to payload. If set pacman downloads to the filename given by the server.
The field trust_remote_name is set in alpm_fetch_pkgurl().
Fixes FS#36791 ([pacman] downloads to wrong filename with redirect).
[dave: remove redundant assignment leading to memory leak]
Signed-off-by: Allan McRae <allan@archlinux.org>
This defines a level of interest a user has in a repository. These are
described by the bitmask flags in the alpm_db_usage_t enum:
ALPM_DB_USAGE_SEARCH: repo is valid for searching
ALPM_DB_USAGE_INSTALL: repo is valid for installs (e.g. -S pkg)
ALPM_DB_USAGE_UPGRADE: repo is valid for sysupgrades
ALPM_DB_USAGE_ALL: all of the above are valid
Explicitly listing the contents of a repo will always be valid, and the
repo will always be refreshed appropriately on sync operations.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
The symbol 'err' refers to err() from err.h, and is wisest to be avoided
as a variable name.
Reference: http://man7.org/linux/man-pages/man3/err.3.html
Signed-off-by: Sami Kerola <kerolasa@iki.fi>
Signed-off-by: Allan McRae <allan@archlinux.org>
* append "/" to directories before searching package file lists
* use lstat over stat so symlinks aren't resolved
* fix the inverted check for stat's return value
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
These references to bug numbers assume we will forever be using that bug
tracker. It is better to properly comment the code instead (which was
done in almost all cases anyway).
Signed-off-by: Allan McRae <allan@archlinux.org>
Using setlocale in the backend is bound to lead to frontend issues
and we have have been using epoch in our databases since April 2007
(commit 47622eef). Remove support for old style times.
Signed-off-by: Allan McRae <allan@archlinux.org>
We currently only warn if a directory's permissions differ, but using -Qkk
on my system shows that directory permissions tend to change in packages
reasonably frequently without notice. Provide a warning in such cases
so that it can be altered. Example output:
(1/1) reinstalling nginx
warning: directory ownership differs on /var/lib/nginx/proxy/
filesystem: 33:0 package: 0:0
Signed-off-by: Allan McRae <allan@archlinux.org>
If a sync DB is malformed and contains entries in the root of the
archive, load_pkg_for_entry will leave the 'filename' variable empty,
leading to a crash in the ensuing strcmp() calls which determine the DB
fragment being examined.
While this isn't a read error, this should be reported to the user so
that it can be addressed by the creator of the DB.
As seen: https://bbs.archlinux.org/viewtopic.php?pid=1297766
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
On operating systems we support, the behavior is always such that the
kernel will do the right thing as far as invalidating the file
descriptor, regardless of the eventual return value. Therefore,
potentially looping and calling close multiple times is wrong.
At best, we call close again on an invalid FD and throw a spurious EBADF
error. At worst, we might close an FD which doesn't belong to us when a
multi-threaded application opens its own file descriptor between
iterations of the loop.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
It is now possible to invert patterns in NoExtract and NoUpgrade.
This feature allows users to whitelist certain files that were
previously blacklisted by another entry.
Signed-off-by: Allan McRae <allan@archlinux.org>
On upgrades, indirect dependencies were not being detected if there was
a dependency in between them that was not part of the transaction. For
example, with the dependency chain: pkg1 -> pkg2 -> pkg3, if pkg1 and
pkg3 are being upgraded but not pkg2 pacman would not order pkg1 and
pkg3 properly.
This was particularly problematic when replacements were involved
because the replaced package(s) would be removed at the start of the
transaction. If an install script required the replacer and lacked
a direct dependency, it could fail.
Fixes FS#32764.
Partially fixes FS#23011.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Dan McGee