Add support for overriding configuration in /etc/makepkg.conf and
~/.makepkg.conf by setting the environment variable PACKAGER similar to
how SRCDEST and PKGDEST behave.
Signed-off-by: Allan McRae <allan@archlinux.org>
This addresses a short but sweet race condition currently existing in
repo-add and repo-remove. We do the smart thing and zip the database to
a location in a temporary directory and not over the original database
directly. However, we then proceed to move this file directly from the
temporary directory to our final location, which is more than likely a
cross-filesystem move (/tmp on tmpfs) and thus non-atomic.
Instead, zip the file to the same directory, prefixing the filename with
'.tmp.'. We then move the file into place. This move is guaranteed to be
atomic, so any reader of the database file will get either the old
version, the new version, or ENOENT.
We also perform a hardlink if possible instead of a move when shifting
the old database out of the way to '.old'; this ensures there is no
chance of a database file not existing during the whole process.
Only one small race condition should now be present- when the database
has been fully moved into place and the signature has not, you may see a
mismatch. There seems to be no good way to address this, and it existed
before this patch.
A final note- if someone had locked-down permissions on the directory
that the database files are in (e.g., could only write to foo.db.tar.gz,
foo.db, foo.db.tar.gz.old, foo.db.old, and the lock file), this would
break.
Signed-off-by: Dan McGee <dan@archlinux.org>
Given our semi-frequent use of pushd/popd, if we are in any directory
but the original and the database path given was relative, we won't
unlock the database file when cleaning up after an error.
Signed-off-by: Dan McGee <dan@archlinux.org>
* Add last-minute changes to NEWS
* Don't treat '_' or '_n' special in scripts when finding translatable
strings; this breaks with one use of `read` and a dummy _ variable
Signed-off-by: Dan McGee <dan@archlinux.org>
This applies to pkgver, pkgrel, and epoch and ensures that any trailing
whitespace outside of the context of the variable declaration itself is
properly trimmed. The Bash parser will ignore this, and so should we.
We don't need to worry about leading space because it would force a
syntax error, or fail validation.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Bump the version, update the translation template files, and fill in
NEWS with relevant commits and changes since 4.0.0.
Signed-off-by: Dan McGee <dan@archlinux.org>
This prevent bsdtar from exploding when install= or changelog= are
present without a value.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Detached sgnature files with extension .sign are accepted by gnupg.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
The default is supposidely 30 seconds from the gpg manpage, but that
sure wasn't what I was seeing- it was somewhere closer to two minutes of
silence. Add a more reasonable 10 second timeout value which should be
good enough for any keyserver that doesn't totally stink at it's job.
Signed-off-by: Dan McGee <dan@archlinux.org>
Replace "/tmp" with "${TMPDIR:-/tmp}" to allow for overriding the
hardcoded path.
Since we only use "/tmp" in conjunction with mktemp(1), we could also
have used "--tmpdir", which is GNU-ish, however (and the BSD counterpart
"-t" has been deprecated in GNU mktemp).
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Dan McGee <dan@archlinux.org>
This includes some fixes to the messages that are displayed when a
signal is caught in makepkg or repo-add:
* Instead of always showing "==> ERROR: TERM signal caught. Exiting...",
replace "TERM" by whatever signal is actually caught.
* Fix a typo in the SIGERR error message in repo-add ("occurred" instead
of "occured"). Francois already fixed this for makepkg in 1e51b81c.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Dan McGee <dan@archlinux.org>
There is a small chance that a user sends SIGINT (or any other signal
that is trapped) when we're already in clean_up() which used to lead to
trap_exit() being executed and the remaining code in clean_up() being
skipped due to the bash signal/trap handler blocking EXIT (since its
handler is already being executed, even if it's interrupted).
In practice, this behaviour caused unexpected results (primarily because
pressing ^C at the wrong time left a lock file behind):
$ ./repo-add extra.db.tar.gz foobar
==> Extracting database to a temporary location...
^C
==> ERROR: Aborted by user! Exiting...
$ ./repo-add extra.db.tar.gz foobar
==> Extracting database to a temporary location...
==> ERROR: File 'foobar' not found.
==> No packages modified, nothing to do.
^C
==> ERROR: Aborted by user! Exiting...
$ ./repo-add extra.db.tar.gz foobar
==> ERROR: Failed to acquire lockfile: extra.db.tar.gz.lck.
==> ERROR: Held by process 18522
Fix this and reduce the chance of race conditions in signal handlers by:
* Unhooking all traps in both clean_up() and trap_exit().
* Call clean_up() explicitly in trap_exit() to make sure we remove the
lock file and the temporary directory even if we send SIGINT when
clean_up() is already being executed but didn't reach the unhook code
yet.
Also, add an optional parameter to clean_up() to allow for setting an
explicit exit code when we call clean_up() from trap_exit().
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Dan McGee <dan@archlinux.org>
We have a few incomplete translations, but these should be addressable
before the 4.0.1 maint release that is surely not that far in the
future.
Signed-off-by: Dan McGee <dan@archlinux.org>
Inline comments after pkgver or pkgrel would cause the sanity
checks to fail so remove them before checking the value.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows it to serve double-duty. In order to allow users to base
verification decisions off of both a valid signature and a trusted
signature, we need to assign some level of owner trust to the keys we
designate as trusted on import.
Signed-off-by: Dan McGee <dan@archlinux.org>
* --import now only imports keys from pubkey.gpg and does not import
owner trust; if you want to have both simply run the operations in
sequence.
* --import-trustdb has been simplified; it will overwrite existing
values in the trust database as before, but there is no need to export
it first as those values are safe if left untouched.
* Fix the manpage referring to a non-existent option.
Signed-off-by: Dan McGee <dan@archlinux.org>
If PKGEXT is not one of the recognized tar*'s, create_package() would
create an empty package file and fail, since bsdtar on the left side of
the pipe returns 141 on SIGPIPE (broken pipe).
This patch changes the behavior for an invalid PKGEXT. A warning is
printed on stderr, and a tar file is created. Also retire the obsolete
$EXT variable.
Add the obligatory comment why we don't use bsdtar's compression.
Finally, fix mixed-tab-space indentation.
Signed-off-by: lolilolicon <lolilolicon@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
We're ill equipped to be using this flag as we don't trap and respond to
the ERR signal. The result is that if is ever tripped, pacman-key will
instantly exit with no indication of why. At the same time, we're
already fairly good about doing our own error checking and verbalizing
it before dying.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This prevents the error trap being set off when GPGDir is commented
in pacman.conf. Bug introduced in 507b01b9.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Always quote the right-hand side of expression when the == or != operator
is used, unless intended as a pattern.
Signed-off-by: lolilolicon <lolilolicon@gmail.com>
Always quote the righthand side of expression when the == or != operator
is used, unless intended as a pattern. Quoting bash(1):
When the == and != operators are used, the string to the right of the
operator is considered a pattern. Any part of the pattern may be quoted
to force it to be matched as a string.
Signed-off-by: lolilolicon <lolilolicon@gmail.com>
If '-' isn't the last item, it's interpreted as a range and not
literally, causing problematic behavior in parsing optdepends.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Keep the non-zero return val to let the caller know that the key wasn't
found.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Many PKGBUILDs use formatting whitespace when specifying optdepends.
This is removed when adding a package to a repo-database so the
output of "pacman -Si <package>" and "pacman -Qip <package file>"
becomes inconsistent. Instead, do the adjustment when creating
the .PKGINFO file.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Allow the specification of versioned optdepends with an epoch.
This also (partially) enforces a whitespace between ":" and the
description which is required for the future optdepends parsing
code.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This is similar to the 'foo-revoked' file we had. This will be used to
inform the user what keys in the shipped keyring need to be explicitly
trusted by the user.
A distro such as Arch will likely have 3-4 master keys listed in this
trusted file, but an additional 25 developer keys present in the keyring
that the user shouldn't have to directly sign.
We use this list to prompt the user to sign the keys locally. If the key
is already signed locally gpg will print a bit of junk but will continue
without pestering the user.
Signed-off-by: Dan McGee <dan@archlinux.org>
We're putting the cart ahead of the horse a bit here. Given that our
keyring is not one where everything is implicitly trusted (ala gpgv),
keeping or deleting a key has no bearing on its trusted status, only
whether we can actually verify things signed by said key.
If we need to address this down the road, we can find a solution that
works for the problem at hand rather than trying to solve it now before
signing is even widespread.
Signed-off-by: Dan McGee <dan@archlinux.org>
Unlike our protégé apt-key, removing a key from our keyring is not
sufficient to prevent it from being trusted or used for verification. We
are better off flagging it as disabled and leaving it in the keyring so
it cannot be reimported or fetched at a later date from a keyserver and
continue to be used.
Implement the logic to disable the key instead of delete it, figuring
out --command-fd in the process.
Note that the surefire way to disable a key involves including said key
in the keyring package, such that it is both in foobar.gpg and
foobar-revoked.
Signed-off-by: Dan McGee <dan@archlinux.org>
This finishes the cleanup started in 710e83999b. We can do a straight
import from another keyring rather than all the funky parsing and piping
business we were doing.
Signed-off-by: Dan McGee <dan@archlinux.org>
Otherwise we're hiding extremely relevant bits like this one:
gpg: WARNING: This key is not certified with a trusted signature!
gpg: There is no indication that the signature belongs to the owner.
Signed-off-by: Dan McGee <dan@archlinux.org>
* Ensure usage message is indented correctly
* Show short filenames for both the gpg keyring and revocation file
Signed-off-by: Dan McGee <dan@archlinux.org>
This fixes build errors when performing a manual install straight to a
filesystem where the files already exist.
Reported-by: Sergej Pupykin <ml@sergej.pp.ru>
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
4ed12ae tightened up the logic to use only find, but ignored the fact
that since the manpage hard link names were no longer captured. They
were created as separate compressed manpages, rather than as hardlinks.
This also introduces a minor efficiency of deleting all hardlinks at
once and using proper iteration over an array rather than a string.
Note to anyone else touching this code: e2fsprogs and libpcap are useful
for testing this. If that changes in the future, you can use the below
bash to locate others:
IFS=$'\n' read -rd '' -a a < <(find /usr/share/man -type f \! -links 1)
pacman -Qqo "${a[@]}" | sort -u
I broke it!
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
When a sourceball passes this check without any warnings, a newline is
omitted. Similar to the if clause of this else block, print a single new
line at the end of the clause instead of accounting for each output.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
This prevents makepkg from aborting with 'file not found' when
changelog= or install= are declared in a PKGBUILD, but empty.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
This is a fix for a bash3 specific bug, where a file sourced by
/etc/profile would exit non-zero and make its way back up to makepkg,
forcing it to exit after package installation. Along with unsetting the
ERR handler, temporarily unset errexit to avoid this.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
- display associated warnings on same line as pass/fail msg, to be more
consistent with checksum verification output
- properly error on a revoked key (matching pacman's behavior)
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
We seem to enjoy using bash regex capabilities, but never referencing
the result with BASH_REMATCH. Replace almost all regexes with equivalent
globs which are faster and functionally equivalent in these cases.
This enables the extglob shopt.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Don't just set the flag variables to zero length strings, actually unset
them from the environment. This fixes issues with broken gnu Makefies
that use ?= for assigment, where the presence of a var is enough to make
this condition avoid assignment.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
- Properly read each sorted line into a new array, instead of breaking
on every word.
- LC_COLLATE should apply to the sort portion of the pipeline, not the
printing.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
This also renames '--receive' to '-recv-keys' to match the wrapped gpg
option name, rather than invent a new one, now that the calling
convention is the same.
Signed-off-by: Dan McGee <dan@archlinux.org>
We were using the mystical [<foobar>] options which is some sort of
cross between a <required> argument and an [optional] one. Remove this
madness and do some other general cleanup/consistency work in the
manpage.
Signed-off-by: Dan McGee <dan@archlinux.org>
The regex wasn't rooted at the end of the filename, nor was it matching
a period/dot before the file extension. The end result was this matched a
file named '07_all_sig.patch' which is totally broken.
Signed-off-by: Dan McGee <dan@archlinux.org>
Read the entire variable, respecting escapes, which are necessary to
retain for the successive eval.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Further improvments on 2ca27ab which will allow the changelog and
install script files to contain whitespace.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Broken in 2ca27a by me, trying to fix another problem.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows local signing of a given key to help establish the web of
trust rooted at the generated (or imported) master key.
Signed-off-by: Dan McGee <dan@archlinux.org>
This enables pacman-key, during --init, to generate a single secret key
for the pacman keyring if one is not present. This will be used as the
root of the web of trust for those that do not wish to manage it with
their own key, as will be the default.
This does not preclude later adding other secret keys to the keyring, or
removing this one- we simply ensure you have at least one secret key
available.
Signed-off-by: Dan McGee <dan@archlinux.org>
Any option that flips UPDATEDB=1 doesn't work right now due to what we
thought was a good idea in commit cab1379a1a. Fix this by not
including the update operation in the option count and special casing
it where necessary.
Also, bring back the helpful "Updating trust database" message.
Signed-off-by: Dan McGee <dan@archlinux.org>
This adds a add_gpg_conf_option() helper function which tries to be
intelligent and only add not found options, and those which have not
been explicitly commented out.
The new options added are 'no-greeting', 'no-permission-warning', and a
default 'keyserver'.
Signed-off-by: Dan McGee <dan@archlinux.org>
* secring.gpg can be 600, readable by root user only
* ensure grep for lock-never option in check_keyring doesn't catch comments
Signed-off-by: Dan McGee <dan@archlinux.org>
The HoldKey option was undocumented and was not suited for pacman.conf.
Instead use the file "/etc/pacman.d/gnupg/heldkeys" to contain a list
of keys not to be removed from the pacman keyring with the --populate
option.
Signed-off-by: Allan McRae <allan@archlinux.org>
After most operations that touch the keyring, it is a good idea to
always run a check on the trustdb as this prevents gpg complaining
on later operations.
Inspiration-from: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
The optimization of only importing keys that were not to be later
revoked was a not smart enough. For example, if a key was
in both a repos keyring and its revoke list, alternate runs of
pacman-key --populate would add then remove the key from the pacman
keyring. This problem is made worse when considering the possibility
of multiple keyrings being imported.
Instead, import all keys followed by the revoking of all keys. This
may result in a key being added then revoked, but that is not much of
an issue given that is a very fast operation.
Signed-off-by: Allan McRae <allan@archlinux.org>
The current --reload option, apart from being non-clear in its naming,
is very limited in that only one keyring can be provided. A distribution
may want to provide multiple keyrings for various subsets of its
organisation or custom repo providers may also want to provide a keyring.
This patch adds a --populate option that reads keyrings from (by default)
/usr/share/pacman/keyrings. A keyring is named foo.gpg, with optional
foo-revoked file providing a list of revoked key ids. These files are
required to be signed (detached) by a key trusted by pacman-key, in
practice probably by the key that signed the package providing these
files. The --populate flag either updates the pacman keyring using all
keyrings in the directory or individual keyrings can be specified.
Signed-off-by: Allan McRae <allan@archlinux.org>
This is a safety measure to prevent simple code injection.
$ i="foo bar"
$ eval i="$i"
bash: bar: command not found
$ eval i=\"$i\"
$ echo "|$i|"
|foo bar|
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
If any of pkgver, pkgrel or epoch contained a variable substitution,
then it needed to be evaluated before checking its value conformed
to the rules.
[Dan: add quotes around RHS]
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
makepkg --source is a often used go make source package like for AUR.
Have a -S shortcut will save the world.
Signed-off-by: Sebastien Luttringer <seblu@seblu.net>
Signed-off-by: Dan McGee <dan@archlinux.org>
When creating a repo outside the current directory, the signature
symlink was not created.
Reported-by: Gaetan Bisson <bisson@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
The use of warning once we had already started adding a package was
confusing as it broke the standard indent pattern. It was especially bad
if adding multiple packages as it wasn't clear what sub-messages applied
to which package being added. This should be an output change only from:
==> Adding package '/tmp/sync/netcfg-2.6.7-1-any.pkg.tar.xz'
-> Computing checksums...
-> Adding package signature...
==> WARNING: An entry for 'netcfg-2.6.7-1' already existed
-> Removing existing entry 'netcfg-2.6.7-1'...
-> Creating 'desc' db entry...
-> Creating 'depends' db entry...
to:
==> Adding package '/tmp/sync/netcfg-2.6.7-1-any.pkg.tar.xz'
==> WARNING: An entry for 'netcfg-2.6.7-1' already existed
-> Computing checksums...
-> Adding package signature...
-> Removing existing entry 'netcfg-2.6.7-1'...
-> Creating 'desc' db entry...
-> Creating 'depends' db entry...
Signed-off-by: Dan McGee <dan@archlinux.org>
These are equivalent. Use the autoconf macro for consistency.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This applies to the repo-remove man page as well as the script itself.
Yes Dan, I ran distcheck afterwards.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
When adding a package to a repo, it is useful to be able to see
that repo-add has indeed found the signature file.
[Dan: update text to be more in line with other messages]
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
