Refactor the upgrade routine to use an array that can be allocated in
one step instead of an alpm_list that is gradually extended in loops.
Signed-off-by: Sören Brinkmann <soeren.brinkmann@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Check the return value of malloc() before dereferencing the returned pointer.
Signed-off-by: Sören Brinkmann <soeren.brinkmann@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Packaging outside of fakeroot can result in incorrect permissions for
package files. It has been years since fakeroot issues during packaging
were common, so it is safe to enforce fakeroot usage. If using fakeroot
is impossible for some reason, a two line wrapper script will suffice to
fool makepkg.
Signed-off-by: Allan McRae <allan@archlinux.org>
The days of fakeroot being error ridden are long gone, so there is no
valid reason to run makepkg as root.
Signed-off-by: Allan McRae <allan@archlinux.org>
1. Packagers who want to test the package() function, i.e. to check the content
of the pkg/ folder.
2. Developers who want to check how the packaged version of a program looks, in
other words how the pkg/ folder looks.
3. For users of systems with no port tree, makepkg can ease package creation.
However the resulting archive of the whole makepkg process is often useless.
For all situations, makepkg will usually be called several times. But no archive
(the final package) is needed in any cases. The archive creation ends up being a
waste of time and resource, especially for big applications and slow machines.
Since this option aborts the process prematurely, it behaves like the
-o,--nobuild option, i.e. any other option acting on later stages in the process
will be automatically discarded. For --noarchive, it means that in
$ makepkg --noarchive --install
the '--install' option does not do anything.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Previously, install_package() return code was either used as exit code or
ignored. This was rather confusing. The return code is now always ignored and a
comment on install_package() has been added.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Another small patch that I made long ago (but was too lazy to submit
back then).
When using XferCommand, TotalDownload option is not take into account.
Regards, Matthias
>From bff946cd9969b6bc6243875e0759f0ee02e487de Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Matthias=20Kr=C3=BCger?= <matthias.krueger@famsik.de>
Date: Thu, 9 May 2013 13:21:04 +0200
Subject: [PATCH] doc: pacman.conf: note that TotalDownload and XferCommand
options are incompatible.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Signed-off-by: Matthias Krüger <matthias.krueger@famsik.de>
Signed-off-by: Allan McRae <allan@archlinux.org>
Use MALLOC instead of malloc for safety in libalpm. Some changes are pure
refactoring, but for others this provides a success check for memory
allocation.
Signed-off-by: Allan McRae <allan@archlinux.org>
This makes cleaning up /tmp after several --keep-root runs much easier.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Files hosted on kernel.org only provide signatures for the uncompressed tarball.
Support this scheme by transparently uncompressing the archives and piping
the data into gpg.
Signed-off-by: Allan McRae <allan@archlinux.org>
Previously only one pattern was allowed.
$ pacsearch foo bar
Search for packages containing 'foo bar'.
$ pacman -Ss foo bar
Search for packages containing both 'foo' and 'bar'.
Note that removing the quotes from the call was not enough since
$ pacsearch 'foo|bar'
would then fail.
Note the use of '--' to indicate the end of option parsing. This way we ensure
that input will always be valid and we need not input checks anymore.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
ALPM still adds a warning to the log, but doesn't emit an event about
said warning, instead using a specific event to let the frontend what
happened/how to inform the user.
Note that there are 2 cases for installing a .pacnew file, to not
overwrite user changes and because file is in NoUpgrade. In the later case
the warning was a bit different: it happened before and said "extracting"
instead of "installed." Now both happen after and are phrased the same.
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
We include the leading space in the match for 'group' and 'installed'. This
allows us to remove the conditions when printing.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Package are processed in the same order as pacman output, so there is no real
need to sort. This makes the code simpler and faster.
The only difference is that local packages will always be printed at the
end. Previously, they were printed before multilib for instance.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
In the old pacsearch, packages were identified uniquely by pkgfields[1], which
contained pkgname+pkgver. Since commit 4d13558 pkgver is stored in pkgfields[2],
and packages have been identified with pkgfields[1] only. Because of that
packages with a different version would appear once only.
This fixes the regression by identifying packages with both pkgfields[1] and
pkgfields[2].
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
I noticed when I built a kernel package with options 'strip' and 'debug' that
the debug package was missing the debug information for the kernel modules.
This patch adds kernel modules to the targets for 'strip' and is compatible
with the split-out debug information packages.
v2: Correct logic error -- did a 'continue' after the 2nd-level case statement
instead of in the default case.
v3: Oops, forgot to terminate outer case. Ran this version through the test
suite and all tests passed.
Signed-off-by: Steven Noonan <steven@uplinklabs.net>
Signed-off-by: Allan McRae <allan@archlinux.org>
This moves the "wrong args" error up from trans_commit to add_pkg when
used with a local pkg and adds the error for remove_pkg when used with
a sync pkg, which currently just removes the db entry.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
These will be emmitted when download a package file from a repository,
indicating that the download starts, and whether it was successfull or
not.
Note that when multiple servers are available, no event is emmitted when
switching to another server.
(This doesn't apply to alpm_fetch_pkgurl(), but since it is called by
the frontend, it shouldn't have problems knowing when the download
starts and when it ends.)
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
To go along with RETRIEVE_START, one other event will be emmitted once
the downloads are done: RETRIEVE_DONE if all files were successfully
downloaded, else RETRIEVE_FAILED.
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
When ALPM emits a log message, it still goes through _alpm_log() but
instead of calling a specific log callback, it goes as an event.
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Instead of using two void* arguments for all events, we now send one
pointer to an alpm_event_t struct. This contains the type of event that
was triggered.
With this information, the pointer can then be typecasted to the
event-specific struct in order to get additional arguments.
Signed-off-by: Olivier Brunel <jjk@jjacky.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
This greatly speeds up file list generation times by avoiding
uncompressing the whole package.
pacman -S base with a deliberate file conflict:
before: 9.1 seconds
after: 2.2 seconds
Signed-off-by: Florian Pritz <bluewind@xinu.at>
If the user opted not to remove the unresolvable packages from the
transaction, the list was neither free'd nor saved to the transaction to
be free'd in trans_release.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Linear output is especially useful when piped to other commands, e.g.
$ pactree -lu pacman | pacman -Qqkk -
The above command would previously show errors on packages with provisions.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Front-ends should be able to free memory that alpm hands them.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
Succeeds if the specified path is a file and is empty.
Signed-off-by: Andrew Gregory <andrew.gregory.8@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
makepkg only aborts on errors during PKGBUILD functions so the remaining
"|| true" statements are unneeded.
Signed-off-by: Allan McRae <allan@archlinux.org>
It turns out that you should refer to variables by their actual name...
Introduced in commit a521cea9.
FS#38414.
Signed-off-by: Allan McRae <allan@archlinux.org>
FS#34240. This is useful for translators and has little cost to us since
we moved to transifex for translations.
Signed-off-by: Allan McRae <allan@archlinux.org>
If the user has "armor" in thier gpg.conf, makepkg will create an
ascii-armored signature. Use --no-armor in the gpg call to avoid.
FS#38503.
Signed-off-by: Allan McRae <allan@archlinux.org>
No more per-repo coloring: this was not Arch-agnostic, and there is no
reasonable, simple way to color repos in a consistant manner with only 6 colors.
'local' is in red: this way we benefit from the pacman -Ss && pacman -Qs combo.
to_color subroutine: it takes an array instead of a string, this is faster and
simpler.
Signed-off-by: Pierre Neidhardt <ambrevar@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
