Begin enforcing the need to pass a handle. This allows us to remove one
more extern handle declaration from the backend.
Signed-off-by: Dan McGee <dan@archlinux.org>
This requires a lot of line changes, but not many functional changes as
more often than not our handle variable is already available in some
fashion.
Signed-off-by: Dan McGee <dan@archlinux.org>
The few remaining instances were utilized for buffers in calls to
snprintf() and realpath(). Both of these functions will always ensure
the returned value is padded with '\0', so there is no need for the
extra byte.
Signed-off-by: Dan McGee <dan@archlinux.org>
The vast majority of the time we will just be passing the same string
value on to the lstat() call. The only time we need to duplicate it is
if the path ends in '/'. In one run using a profiler, only 400 of the
200,000 calls (0.2%) required the string to be copied first.
Signed-off-by: Dan McGee <dan@archlinux.org>
Due to the way we set up the graph structure, we don't always have good
parent information. The changes made in dd8cf0c12d assumed this, so
back them out and just live with the dead pointers being there in the
memory while we are cleaning up after ourselves.
Signed-off-by: Dan McGee <dan@archlinux.org>
These new method signatures return and take handle objects to operate on
so we can move away from the idea of one global handle in the API. There
is also another important change and that deals with the setting of root
and dbpaths. These are now done at initialization time instead of using
setter methods. This allows the library to operate more safely knowing
that paths won't change underneath it.
Signed-off-by: Dan McGee <dan@archlinux.org>
This keeps duplicate code to a minimum. This will come in more handy as
we refactor some of these option setters away.
Signed-off-by: Dan McGee <dan@archlinux.org>
When only downloading a package, pacman can produce some incorrect
output.
> pacman -Sddw nvidia-utils
warning: nvidia-utils-270.41.19-1 is up to date -- reinstalling
This line is now now silenced when using -Sw.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows us to not require the context (e.g. handle) when calling
this function. Also beef up the checks in the two callers of this
function to bail if the last return code is not ARCHIVE_EOF, which is
the expected value.
This requires a change to one of the pactest return codes and the
overall result of the test, but results in a much safer operating
condition whereby invalid database entries will stop the operation.
Signed-off-by: Dan McGee <dan@archlinux.org>
This kills a lot more global handle business off. sync.c still requires
the handle declaration for one reference that can't be changed yet; it
will be removed in a future patch which isolates all of the necesary API
changes.
Signed-off-by: Dan McGee <dan@archlinux.org>
This will make the patching process less invasive as we start to remove
this variable from all source files.
Signed-off-by: Dan McGee <dan@archlinux.org>
Similar to what we just did for the database; this will make it easy to
always know what handle a given package originated from.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is the first step in a long process to remove our dependence on the
global handle variable we currently share in libalpm, with the goal to
make things a bit more thread-safe and re-entrant.
Signed-off-by: Dan McGee <dan@archlinux.org>
These are simple accessor functions for a struct; the handle never even
comes into play when calling these functions.
Signed-off-by: Dan McGee <dan@archlinux.org>
The usefulness of this is rather limited due to it not being compiled
into production builds. When you do choose to see the output, it is
often overwhelming and not helpful. The best bet is to use a debugger
and/or well-placed fprintf() statements.
Signed-off-by: Dan McGee <dan@archlinux.org>
Callers to curl_download_internal now tell us if its okay to continue a
transfer, so obey this instead of using a heuristic.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
If a connection drops below 1kb/s for 10s, curl will kill the transfer
and we'll report failure. This is the average transfer speed over the
delta defined by CURLOPT_LOW_SPEED_TIME, so setting a low value here
shouldn't bother folks using 14.4k dial-up.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This addresses FS#24292. If one does the bad thing of not checking
pm_errno after calling set_dbpath(), you may not realize the
initialization process went wrong and calling trans_init() resulted in a
segfault. If we don't have a lockfile path, bail out and have
trans_init() fail.
Also remove a ALPM_LOG_FUNC call that was causing pm_errno to return "no
handle"; this was due to a log call in the handle setup (whereby the log
attempts to use a callback attached to the handle).
Signed-off-by: Dan McGee <dan@archlinux.org>
Issue FS#24230. If a symlink is broken and included in the removal
process of a package, we blew up and segfaulted due to
alpm_compute_md5sum() returning NULL and then performing a strcmp()
operation.
Signed-off-by: Dan McGee <dan@archlinux.org>
The addition of the DB version check introduces a lag time between the
lockfile creation and the transaction initialization. In cases where the
local DB is large enough and/or the user's disk is slow enough, this
time is significant enough that its possible for a user to send a SIGINT
and leave behind a db.lck file.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
It's your own damn fault if you do this, and this code is remnants from
an old time when we weren't very good at coding.
Signed-off-by: Dan McGee <dan@archlinux.org>
This needlessly prevents the easiest way available of clearing any of these
values. We can also do the same for the 'arch' value.
Signed-off-by: Dan McGee <dan@archlinux.org>
Currently we have one call that has all sorts of crazy behavior and doesn't
make a whole lot of sense. Go from one method to the normal four methods we
have for all of our other lists we use in the library to make it a lot
easier for a frontend to manipulate server lists.
Signed-off-by: Dan McGee <dan@archlinux.org>
The switch from FUNCTION to DEBUG was ill-advised inside the local
database load. Instead, add a DEBUG level logger to both local and sync
database loads that shows the number of packages processed.
Signed-off-by: Dan McGee <dan@archlinux.org>
We were erroring out in the case where a first (possibly bogus) mirror
would cause the download process to return a failure code, even though
subsequent servers had the file.
Signed-off-by: Dan McGee <dan@archlinux.org>
This started off removing the "(void)foo" hacks to work around
unused function parameters and ended up fixing every warning
generated by -Wunused-parameter.
Dan: rename to UNUSED.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
We had a lot of similar looking code that we can collapse down into a
function. This also fixes errors seen when turning on some gcc warnings
and implicitly casting away the const-ness of the string. Free the list
when we are done with it as well.
Also, fix a logic error where we should be checking with &&, not ||.
Signed-off-by: Dan McGee <dan@archlinux.org>
This makes it possible to omit usage of -lgpgme, just as we can do for
-lcurl and -lcrypto.
Thanks to Rémy Oudompheng for an initial stab at this.
Signed-off-by: Dan McGee <dan@archlinux.org>
Add some lookup functions for nice names for the various types used by
the library, and remove some fields that are of little use to us in the
debug output. This should make looking at key loading and verification a
bit easier, especially in determining what makes up our good and bad
criteria.
Signed-off-by: Dan McGee <dan@archlinux.org>
Both md5sum verification and PGP verification can and should be done at
package load time. This allows verification to happen as early as
possible for packages provided by filename and loaded in the frontend,
and moves more stuff out of sync_commit that doesn't really belong
there. This should also set the stage for simplified parallel loading of
packages later down the road.
Signed-off-by: Dan McGee <dan@archlinux.org>
More stuff going on in the pre-committing stage that can be in a static
method to make things a bit more clear.
Signed-off-by: Dan McGee <dan@archlinux.org>
This part is almost completely self-contained, except building the list
of delta filenames that we use later to check their md5sums. Refactor it
into a static method so we can bring most of the code in sync_commit
closer to the method name.
Signed-off-by: Dan McGee <dan@archlinux.org>
Given that we offer no transparency into the pmpgpsig_t type, we don't
really need to expose it outside of the library, and at this point, we
don't need it at all. Don't decode anything except when checking
signatures. For packages/files not from a sync database, we now just
read the signature file directly anyway.
Also push the decoding logic down further into the check method so we
don't need this hanging out in a less than ideal place. This will make
it easier to conditionally compile things down the road.
Signed-off-by: Dan McGee <dan@archlinux.org>
Rather than go through all the hassle of doing this ourselves, just let
GPGME handle the work by passing it a file handle.
Signed-off-by: Dan McGee <dan@archlinux.org>
The various "level" values were a bit crazy to decipher, and we were
doing some very interesting comparisons in certain places. Break it out
into two parameters instead so we can seperate the type from the extra
information display, and do things accordingly.
Nothing changes with the display of any of the five types we currently
show: -Si, -Sii, -Qi, -Qii, -Qip.
Something to note- we should expose the PKG_FROM enum type somehow, this
patch leaves the door open to do that quite easily.
Signed-off-by: Dan McGee <dan@archlinux.org>
There's a lot of related moving parts here:
* Iteration through mirrors is moved back to the calling functions. This
allows removal of _alpm_download_single_file and _alpm_download_files.
* The download function gets a few more arguments to influence behavior.
This allows several different scenarios to customize behavior:
- database
- database signature (req'd and optional)
- package
- package via direct URL
- package signature via direct URL (req'd and optional)
* For databases, we need signatures from the same mirror, so structure
the code accordingly.
Some-inspiration-from: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
The allow_resume is the start of the fix to the "don't ever resume
database downloads" problem, as well as being useful for '.sig'
downloads as well. For now, we say "always allow resume", but this will
eventually get pushed down as necessary.
Error checks are reworked in order to correctly error out when a file is
not found on the remote end and reports 0 bytes downloaded. In addition,
the two error messages printed are now different as one reports a more
specific error message provided via the cURL error buffer.
Some example output from an -Sy run with [testing], [community],
[community2], [eee], and [nonexistant] defined as repos. [community2]
and [nonexistant] are both invalid, one using FTP and one using HTTP.
:: Synchronizing package databases...
testing is up to date
community is up to date
error: failed retrieving file 'community2.db' from ftp.archlinux.org : Given file does not exist
error: failed to update community2 (FTP: couldn't retrieve (RETR failed) the specified file)
eee is up to date
error: failed retrieving file 'nonexistant.db' from code.toofishes.net : The requested URL returned error: 404
error: failed to update nonexistant (HTTP response code said error)
Signed-off-by: Dan McGee <dan@archlinux.org>
The value PM_PGP_VERIFY_UNKNOWN is reserved to error cases,
now that the signature verification level defaults to the
globally set level. The only error case is when handle == NULL,
which is false in the context of _alpm_sync_commit().
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This does touch a lot of things, and hopefully doesn't break things on
other platforms, but allows us to also clean up a bunch of crud that no
longer needs to be there.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is the standard, and we have had a few of these introduced lately
that should not be here.
Done with:
find -name '*.c' | xargs sed -i -e 's#if (#if(#g'
find -name '*.c' | xargs sed -i -e 's#while (#while(#g'
Signed-off-by: Dan McGee <dan@archlinux.org>
* add _alpm_db_get_sigverify_level
* add alpm_option_{get,set}_default_sigverify
And set the default verification level to OPTIONAL if not set otherwise.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This makes it absolutely dead easy to ensure off_t has the same length
in all compilation units. I just spent 2.5 hours bashing my head on an
issue related to this so damn it I'm fixing it for good.
Signed-off-by: Dan McGee <dan@archlinux.org>
Currently the only error case then when handle == NULL.
However several handle functions return -1 on this error,
and a uniform API makes things simpler.
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
For a package to be loaded from any of our backends, these two fields
are always required upfront. Due to this fact, we don't need them to be
backend-specific operations and can just refer to the field directly.
Additionally, our static (and thus private) cache package accessors had
a NULL check on pkg before returning the relevant field. Eliminate this
since they only way they are ever called is via the packages attached
callback struct, which would have caused the NULL pointer dereference in
the first place.
Signed-off-by: Dan McGee <dan@archlinux.org>
So we only need one copy in the final library, not one copy per time
used. Ensure all necessary includes are in place (especially to get the
right size of off_t each time it is compiled) by including "config.h" in
the new graph.c.
One small adjustment here makes the graph_free code more robust- ensure
we don't have invalid pointers after each iteration by looking at the
parents and children and adjusting accordingly.
Signed-off-by: Dan McGee <dan@archlinux.org>
Ensure we only have one- this looks like the result of a bad merge from
old 2008 signing code with the current stuff which has changed quite a
bit.
Originally-seen-by: Rémy Oudompheng <remyoudompheng@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
The current state of the code does not allow to see immediately
that it returns a list of pmdepmissing_t structures.
Signed-off-by: Dan McGee <dan@archlinux.org>
Some systems, like FreeBSD might define both statfs
and statvfs: however if statvfs exists whereas getmntinfo()
uses a statfs struct, the current ifdefs would select the wrong
line of code.
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
There is no reason to not support versions of libarchive that lack
ARCHIVE_COMPRESSION_UU. Distributions should work properly without
this.
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
That's a funny one, building with optimization levels (with both gcc and
clang) caused open_mode to always be set to "ab", which worked.
This was spotted both with clang-analyzer, and by Jakob who reported a
segfault as he was using an un-optimized build.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
So we don't segfault when calling this on be_sync loaded packages. They
return logical values as much as possible for indicating there is no
changelog available.
Signed-off-by: Dan McGee <dan@archlinux.org>
We didn't do this sanity check before trying to open an archive. If
the alpm dbpath wasn't set, the sync database dbpath would be NULL,
causing us to hang indefinitely in archive_read_open_filename() rather
than erroring out.
We already have a corresponding check in local_db_populate().
The following program will test this case, and hangs before this patch
without the call to set_dbpath:
int main(int argc, char *argv[]) {
alpm_initialize();
// alpm_option_set_dbpath("/var/lib/pacman/");
pmdb_t *core = alpm_db_register_sync("core");
pmpkg_t *pkg = alpm_db_get_pkg(core, "pacman");
return 0;
}
Signed-off-by: Dan McGee <dan@archlinux.org>
We had two functions that were oh so similar but slightly different. We
can combine them and add some conditional operation stuff to decide what
to return.
Signed-off-by: Dan McGee <dan@archlinux.org>
Many alpm_option_get/set_*() functions already check this
and set pm_errno to the right value, but not all, so
this improves consistency.
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This greatly simplifies the cleanup fallthrough in our download function
and we'll be able to reuse this for signatures.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Based on the fact that localf always points to the same file, there's no
need to code in multiple fopen calls with varying results. Instead,
track the desired file open mode and make a single call to fopen.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Create a more general function that allows appending a suffix to a
filepath.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This lets us determine the real size of the file on disk so that we can
properly bump the progress bar when we're resuming a download.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This addresses FS#23424. The -dd backend code was introduced in commit
b6ec9019d7, and unfortunately the munged depend used for comparison did
not carry through to the eventual display of this version. To fix this,
we undo some of the depcmp_tolerant() business introduced, and instead
make a new pmdepend_t object if necessary when the no dependency version
flag is set. This results in the correct depend being copied to the
missing depend passed onto the frontend.
Signed-off-by: Dan McGee <dan@archlinux.org>
pacman 3.5.0 removed alpm_db_register_local, so calling
alpm_db_unregister_all leaves the front end in a position where there's
no local db, and no way to re-register it.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
After updating a database, remove the old signature to prevent it
being used in validation if the new signature fails to download.
Signed-off-by: Allan McRae <allan@archlinux.org>
If signature verification is needed, attempt to download a signature
file for a repo when it is updated. Return an error if unable to
download signature only when checking is mandatory, or if signature is
invalid.
TODO: At the moment the database signature is only checked on download.
Should we do anything with a database if it fails to be verified to prevent
its future usage?
Signed-off-by: Allan McRae <allan@archlinux.org>
Add a pmpgpsig_t struct to the database entry struct and functions for
the lazy loading of database signatures. Add a function for checking
database signatures, reusing (and generalizing) the code currently used
for checking package signatures.
TODO: The code for reading in signature files from the filesystem is
duplicated for local packages and database and needs refactoring.
Signed-off-by: Allan McRae <allan@archlinux.org>
Our curl callback does a whole lot of work for nothing if the front end
never defined a callback to receive the data we'd calculate for it.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
CURLINFO_HTTP_CODE is deprecated in favor of CURLINFO_RESPONSE_CODE.
Both yield the same values.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
The files we transfer are generally compressed already, so this just
adds unnecessary overhead.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Use a static variable to effectively track the initialization state of
the progress callback via the last byte amount reported as downloaded by
libcurl.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
* introduces new macro in util.h (DOUBLE_EQ) for properly comparing
floating point values
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Move the (possibly still temporary) output generated during signature
checking into the --debug output.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Once we do this, add support for VerifySig to pactest. We just check if
the repo name contains Always, Never or Optional to determine the value
of VerifySig. The default is Never. pacman uses Always by default but
this is not suitable for pactest.
Original-work-by: shankar <jatheendra@gmail.com>
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
We pass in a db object, so no need to go looking for it in the list on
the handle. This is a remnant of when we passed in a treename, more than
likely.
Signed-off-by: Dan McGee <dan@archlinux.org>
This will serve as the home directory we pass to GPGME when making calls so
we can have a libalpm-utilized keyring.
Signed-off-by: Dan McGee <dan@archlinux.org>
If a .sig file sits side-by-side on the filesystem with a package archive,
read it in during the package struct creation process so we can verify it at
a later time if necessary.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Add a new field to the package struct to hold PGP information and
instruct db_read to pick it up from the database. It is currently unused
internally but this is the first step.
Due to the fact that we store the PGP sig as binary data, we need to store
both the data and the length so we have a small utility struct to assist us.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
We will need these for GPG functionality (decoding the base64 encoded
signature stored in the databases).
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Implements FS#23103. Also modify libalpm so it ignores this value
without any warning as we know it is likely to exist.
Signed-off-by: Dan McGee <dan@archlinux.org>
Wrap lines of long length, noticed while creating and messing around
with some of the other maint branch patches.
Signed-off-by: Dan McGee <dan@archlinux.org>
Noticed with the openoffice/libreoffice replacement scheme where many
packages are listed as replacements to one package, thus electing it for
removal multiple times. Ensure a given package is not already present
before placing it in the removal list.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is a rather serious data corruption issue that luckily manifested
itself today in a noticable way. A package in testing had replaces
entries read in as ["%RE pkgname", "%RE"] which was clearly wrong. This
happens when we hit the end of an archive block, do not have a newline,
and have to continue reading from the next block to complete the line.
Signed-off-by: Dan McGee <dan@archlinux.org>
Drawing progress bars before calling curl_easy_perform() is needless as
the curl progress callback is called with zero progress before actually
downloading the file anyways. Fixes display of "0%" progress bars when
sync'ing package databases that are already up to date.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Dan McGee <dan@archlinux.org>
This was discussed and more or less agreed upon on the mailing list. A
huge checkin, but if we just do it and let people adjust the pain will
end soon enough. Rebasing should be relatively straighforward for anyone
that sees conflicts; just be sure you use the new return style if
possible.
The following semantic patch was used to do the change, along with some
hand-massaging in order to preserve parenthesis where appropriate:
The semantic match that finds this problem is as follows, although some
hand-massaging was done in order to keep parenthesis where appropriate:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression a;
@@
- return(a);
+ return a;
// </smpl>
A macros_file was also provided with the following content:
Additional steps taken, mainly for ASSERT() macros:
$ sed -i -e 's#return(NULL)#return NULL#' lib/libalpm/*.c
$ sed -i -e 's#return(-1)#return -1#' lib/libalpm/*.c
Signed-off-by: Dan McGee <dan@archlinux.org>
We erroniously dropped the call to _alpm_delta_parse() when macro-izing,
causing segfaults for repos that provide deltas. Addresses FS#23314.
Signed-off-by: Dan McGee <dan@archlinux.org>
Partially addresses the "why doesn't CheckSpace work in a chroot" issue.
We can't make it work, but we can at least detect when it won't work by
checking for a partition for our given installation root. If we can't
determine the mountpoint for this, bail out with an error.
Signed-off-by: Dan McGee <dan@archlinux.org>
this is just some debuggery to allow pacman to operate with both fetch
and curl at the same time. use the PACMANDL variable to control which
library is used.
Signed-off-by: Dave Reisner <d@falconindy.com>
This is a feature complete re-implementation of the fetch based internal
downloader, with a few improvements:
* support for SSL
* gzip and deflate compression on HTTP connections
* reuses a single connection over the entire session for lower resource
usage.
Signed-off-by: Dave Reisner <d@falconindy.com>
Add PM_ERR_LIBCURL to error enum and handle case in error.c by returning
curl_easy_strerror() based on the error number carried by the gloabl alpm
handle.
Signed-off-by: Dave Reisner <d@falconindy.com>
no actual code changes here. change preprocessor logic to include
get_tempfile, get_destfile, signal handler enum, and the interrupt
handler logic when either HAVE_LIBCURL or HAVE_LIBFETCH are defined.
Signed-off-by: Dave Reisner <d@falconindy.com>
Do this in preparation for implementing similar curl based
functionality. We want the ability to test these side by side.
Signed-off-by: Dave Reisner <d@falconindy.com>
Adding the CURLcode is necessary in order to return an error string from
pm_error. Unlike libfetch, curl returns numerical error numbers and does
not maintain a staticly allocated string with the last error generated.
Adding the curl object itself to the handle is advantageous (and
encouraged by curl_easy_perform(3)) because the handle is reusable for
successive operations. This cuts back on overhead when downloading
multiple files in a single transaction.
Signed-off-by: Dave Reisner <d@falconindy.com>
repo-add can add a "files" entry into the sync db. Currently we
do nothing with this file, so explicitly skip it to prevent
unknown database file warnings.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Our keywords were all screwed up in this regard. Fix it so our
ngettext() shortcut calls are actually recognized and respected.
Signed-off-by: Dan McGee <dan@archlinux.org>
Fixes FS#23090, a rather serious problem where the user was completely
unable to read the local database. Even if entry->d_type is available,
the given filesystem providing it may not fill the contents, in which
case we should fall back to a stat() as we did before. In this case, the
filesystem was XFS but there may be others.
Signed-off-by: Dan McGee <dan@archlinux.org>
A lot of these were places that should have used the same message but
didn't, or were very easy to convert to using the same message and
letting some of the burden off of the translators.
Signed-off-by: Dan McGee <dan@archlinux.org>
Ensure we have a local DB version that is up to par with what we expect
before we go down any road that might modify it. This should prevent
stupid mistakes with the 3.5.X upgrade and people not running
pacman-db-upgrade after the transaction as they will need to.
Signed-off-by: Dan McGee <dan@archlinux.org>
We only call these from the transaction init and teardown, so move them
to that file, mark them static, and push more of the logic of handle
manipulation into these functions.
Signed-off-by: Dan McGee <dan@archlinux.org>
According to FOPEN(3), using fclose on an fdopen'd file stream also
closes the underlying file descriptor. This happened in _alpm_lckmk
(util.c), which meant that when alpm_trans_release closed it again, the
log file (which reused the original file descriptor) was closed instead.
Signed-off-by: Jonathan Conder <jonno.conder@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
There's no API functions exposed which allow manipulation of this type,
so remove it from public view. Also, rename the public and private
alpm_db_get_pkgcache symbol to alpm_db_get_pkgcache_has.
Signed-off-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows error messages emitted by the frontend to be a bit more
descriptive and not have the annoying "well why didn't you tell me that
the first time" problem. If a package had multiple missing deps, we
would bail on the first one before rather than finish processing all
missing dependencies, and only print one error message. Instead,
continue through this entire set of missing deps and append all eventual
errors.
The added pactest tests this case, as the to be installed package has
two missing dependencies. However, pactest does not actually test or see
the difference in output from before and after, so it passes in both
cases, but it is clearly visible in the logs.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is old code that has since gone stale; we no longer ever add
anything to this list so no need to keep it around and check the
contents during extraction.
Signed-off-by: Dan McGee <dan@archlinux.org>
* Make conflict_isin() static; it is used nowhere else.
* Remove does_conflict(): it turns out to be replaceable by a single call to
_alpm_depcmp(). By pushing it up, we can reduce calls to _alpm_splitdep()
from 60,368 to 16,940 during one test -Su operation I ran.
Signed-off-by: Dan McGee <dan@archlinux.org>
* Use stat() and not lstat(); we don't care for the size of the symlink if
it is one, we want the size of the reference file.
* FS#22896, fix local database estimation on platforms that don't abide by
the nlink assumption for number of children.
* Fix a missing newline on an error message.
Signed-off-by: Dan McGee <dan@archlinux.org>
Change _alpm_graph_new() to use CALLOC to avoid explicit zeroing out of fields
in pmgraph_t.
Signed-off-by: Pang Yan Han <pangyanhan@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Turn it into an enum rather than a boolean, and use a bitmask like we do for
reading DB entries. The relevant flag is turned on in our two calculate
loops, and anything reading the used flag later can decided which flag (or
either) is relevant.
This will allow the read-only partition code to be triggered on a
remove-only operation, e.g. if /boot was read-only and one tried to remove
grub in a sync transaction. Of course, right now, we don't actually run the
diskspace check code in the '-R' codepath.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is a bit of a stopgap solution for the problem, but an easier one than
revamping the file conflict checking code to support the same stuff. Using
some more gross autoconf magic, figure out which struct field we need to
look at to determine read-only status and store that on our mountpoint
struct. If we find out we needed this partition after calculating size
requirements, then toss an error.
Signed-off-by: Dan McGee <dan@archlinux.org>
In the getmntinfo() section, the local variable mnt doesn't exist; this
would have caused a compile error if I had tested the code on such a
platform. Unify both codepaths to just run strlen() on the already copied
mount path instead.
Signed-off-by: Dan McGee <dan@archlinux.org>
In packages, our description file contains:
key = value is here
type entries, and we passed "key " and " value is here" to our strtrim
function, causing us to always memmove the value portion to remove the
space. Since this is a throwaway buffer, do the advancing on our own before
trimming to save the need to shift memory around; "value is here" will now
be passed and strtrim will be responsible for trailing whitespace.
Signed-off-by: Dan McGee <dan@archlinux.org>
We did this in some but not all cases, assuming the 0 value coming out of
libarchive would not be a problem. However, this does not work for "fake"
filesystems such as rpc_pipefs, which reports a free block and total block
count of zero.
Fix this by not ever counting symlinks or directories, and adding a note
explaining that if we someday do count directories, their size needs to be
attributed to the proper place.
This patch also includes a few cleanups/performance tweaks- avoid calling
strlen() on the mountpoint directory string as much by storing this size in
our mountpoint struct, and push the snprintf() call up to the calculate
functions since we were already doing it here in the remove case.
Signed-off-by: Dan McGee <dan@archlinux.org>
The overlapping code in _alpm_pkghash_add() and _alpm_pkghash_add_sorted()
are now in a new static function pkghash_add_pkg(). This function has a
third flag parameter which determines whether the package should be added in
sorted order.
Signed-off-by: Pang Yan Han <pangyanhan@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
In sync_db_populate() and local_db_populate(), a NULL db->pkgcache is not
caught, allowing the functions to continue instead of exiting.
A later alpm_list_msort() call which uses alpm_list_nth() will thus traverse
invalid pointers in a non-existent db->pkgcache->list.
pm_errno is set to PM_ERR_MEMORY as _alpm_pkghash_create() will only return
NULL when we run out of memory / exceed max hash table size. The local/sync
db_populate() functions are also exited.
Signed-off-by: Pang Yan Han <pangyanhan@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
When reading the "desc" file in _alpm_local_db_read(), some
strings are trimmed and checked for length > 0 before their
use/duplication subsequently. They are then trimmed again
when there is no need to.
The following code snippet should illustrate it clearly:
while(fgets(line, sizeof(line), fp) &&
strlen(_alpm_strtrim(line))) {
char *linedup;
STRDUP(linedup, _alpm_strtrim(line), goto error);
info->groups = alpm_list_add(info->groups, linedup);
}
This patch removes the redundant _alpm_strtrim() calls in
_alpm_local_db_read() such as the one inside the STRDUP shown
above.
Signed-off-by: Pang Yan Han <pangyanhan@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
PM_ERR_WRITE is defined in alpm.h but not handled in
alpm_strerror(). This patch corrects that.
Signed-off-by: Pang Yan Han <pangyanhan@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
We located files in a few places but didn't check if they were files or
directories. Ensure they are actually files using stat() and S_ISREG(); this
showed itself when trying to download to the directory name itself in
FS#22645.
Signed-off-by: Dan McGee <dan@archlinux.org>
Rather than potentially move every item to the next NULL, attempt to move at
most one item at a time by iterating backwards from the NULL location in the
hash array. If we move an item, we repeat the process on the now shorter
"chain" until no more items need moving.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This takes in the list and a list item, and does the pointer dance necessary
to remove it from the list regardless of whether it is first, last, or
somewhere in the middle. It is useful for callers that already know what
item needs to be removed and have a pointer to it rather than doing a search
by data that the plain alpm_list_remove() does.
Refactor alpm_list_remove() to use this function as well.
Signed-off-by: Dan McGee <dan@archlinux.org>
Fully removes a package from the hash. Also unify prototype with
removal from an alpm_list_t, fixing issues when removing a package
from the pkgcache.
Signed-off-by: Allan McRae <allan@archlinux.org>
While probably still not optimal in terms of everyday usage in
pacman, this reduces the absolute size increase to "more reasonable"
levels. For databases greater than 5000 in size, the minimum size
increase is used which is still on the order of a 10% increase.
Signed-off-by: Allan McRae <allan@archlinux.org>
Check that the requested size of a pkghash is not beyond the maximum
prime. Also check for successful creation of a new hash before
rehashing.
Signed-off-by: Allan McRae <allan@archlinux.org>
Since the sync database never changes size once we initialize it, we
allow it to be filled a bit more. This reduces the overall memory
footprint needed by the hash table.
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows us to get through the rehash required by smoke001 and pass
all pactests. It is by no means the best or most efficient
implementation but it does do the job.
Signed-off-by: Dan McGee <dan@archlinux.org>
Read the package information for sync/local databases into a pmpkghash_t
structure.
Provide a alpm_db_get_pkgcache_list() method that returns the list from
the hash object. Most usages of alpm_db_get_pkgcache are converted to
this at this stage for ease of implementation. Review whether these are
better accessing the hash table directly at a later stage.
Signed-off-by: Allan McRae <allan@archlinux.org>
This works for both local and sync databases in slightly different ways. For
the local database, we can use the directory hard link count on the local/
folder. For sync databases, we use the archive size coupled with some
computed average per-package sizes to determine an estimate.
This is currently a dead assignment once calculated, but could be used to
set the initial size of a hash table.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Noted in FS#22697. When I factored out _alpm_parsedate() into a common
function, I didn't move the <locale.h> include properly, causing a build
failure when NLS is disabled and this header isn't automatically included
everywhere.
Signed-off-by: Dan McGee <dan@archlinux.org>
This flag allows to disable version checking in dependency resolving
code.
depcmp_tolerant respects the NODEPVERSION flag but we still keep the
original strict depcmp. The idea is to reduce the impact of the
NODEPVERSION flag by using it in fewer places.
I replaced almost all depcmp calls by depcmp_tolerant in deps.c (except
in the public find_satisfier used by deptest / pacman -T), but I kept
depcmp in sync.c and conflict.c
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
It's likely that these interfaces will break sooner or later, now that
pacman no longer uses them.
So better force the two people who use them to migrate their code to the
new add_pkg/remove_pkg interface, which is very easy anyway.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Old interface is marked as deprecated:
int alpm_sync_target(char *target);
int alpm_sync_dbtarget(char *db, char *target);
int alpm_add_target(char *target);
int alpm_remove_target(char *target);
New recommended interface:
int alpm_add_pkg(pmpkg_t *pkg);
int alpm_remove_pkg(pmpkg_t *pkg);
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
For consistency with alpm_add_pkg.
The new recommended interface is alpm_add_pkg / alpm_remove_pkg, all
others interfaces are deprecated.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
This group function is meant to help group handling from frontend : it
scans all dbs, handling ignored packages and duplicate members (the
first repo where a member is found has the priority).
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
This new function is meant to deprecate all existing
sync/add target functions :
int alpm_sync_target(char *target);
int alpm_sync_dbtarget(char *db, char *target);
int alpm_add_target(char *target);
Rather than dropping these 3 interfaces, it might be better to rewrite
them using alpm_add_pkg for now.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
This is a public interface for resolvedep. It looks nicer to expose it
this way rather than through sync_target.
This function can also be helpful for external tools as it should give
good results close to how pacman select a package for satisfying a given
dep.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
If there are multiple providers in one db, pacman used to just stop at
the first one (both during dependency resolution or for pacman -S
'provision' which uses the same code).
This adds a new conversation callback so that the user can choose which
provider to install. By default (user press enter or --noconfirm), the
first provider is still chosen, so for example the behavior of sync402
and 403 is preserved. But at least the user now has the possibility to
make the right choice in a manual run.
If one of the provider is already installed, it is picked for
reinstall/upgrade, so that provision 002/003 pactest now pass.
$ pacman -S community/smtp-server
:: There are 3 providers available for smtp-server:
1) courier-mta 2) esmtp 3) exim
Which one do you want to install?
Enter a number (default=1):
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Perform the cheap struct and string setup of the local DB at handle
initialization time to match the teardown we do when releasing the handle.
If the local DB is not needed, all real initialization is done lazily after
DB paths and other things have been configured anyway.
Signed-off-by: Dan McGee <dan@archlinux.org>
We explicitly place 'pkgbase' (and used to place 'force') fields inside
PKGINFO files, so ignore them silently instead of printing an error for
them. Also make the error message for unknown keys actually contain the key.
Signed-off-by: Dan McGee <dan@archlinux.org>
We were returning a package error code rather than a DB one, and we
would leak the archive memory if the database file didn't exist.
Signed-off-by: Dan McGee <dan@archlinux.org>
Instead, go the same route we have always taken with version-release in
libalpm and treat it all as one piece of information. Makepkg is the only
script that knows about epoch as a distinct value; from there on out we will
parse out the components as necessary.
This makes the code a lot simpler as far as epoch handling goes. The
downside here is that we are tossing some compatibility to the wind;
packages using force will have to be rebuilt with an incremented epoch to
keep their special status.
Signed-off-by: Dan McGee <dan@archlinux.org>
Adapting from RPM, follow the [epoch:]version[-release] syntax. We can also
borrow some of their parsing code for our purposes (thanks!). Add some new
tests to our vercmp shell script tester for epoch comparisons, and then make
the code work with these newfangled epoch specifiers.
Signed-off-by: Dan McGee <dan@archlinux.org>
Since it is the same string. Done with some bash looping and sed magic.
for src in po/*.po; do
echo $src
newtrans=$(grep -A1 "msgid.*$1" $src | tail -n1)
newtrans=${newtrans//\\/\\\\}
echo "$newtrans"
fname=${src##*/}
dest=lib/libalpm/po/$fname
sed -i -e "/msgid.*$1/{N; s/msgstr.*$/$newtrans/}" $dest
done
Signed-off-by: Dan McGee <dan@archlinux.org>
In most (all?) cases, we will process all files for a given sync database
entry sequentially. The code currently does an _alpm_pkg_find() for every
file in the database, but we had the "current" package readily available.
Shift some local variables around a bit to expose this to sync_db_read() and
use it if the package is the correct one.
On my system, this cuts calls to _alpm_pkg_find() from 20,769 to 10,349
calls during a -Qu operation, and results in a ~30% speedup of the same
operation (0.35 sec -> 0.27 sec). This benefit should be apparent anywhere
we read in the full contents of the sync databases.
Signed-off-by: Dan McGee <dan@archlinux.org>
We don't need to create a temporary copy of the string if we are smart with
our pointer manipulation and string copying. This saves a bunch of string
duplication during database parsing, both local and sync.
Signed-off-by: Dan McGee <dan@archlinux.org>
Remove the need for an unconditional string duplication by using pointer
arithmetic instead, and strndup() instead of an unspecified-length strdup().
This should reduce memory churn a fair amount as this is called pretty
frequently during database loads.
Signed-off-by: Dan McGee <dan@archlinux.org>
When installing packages from a file, the integrity check count
stays at (0/x) complete. This ensures it is bumped to (x/x) at
the end of the process.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is essentially a backport/cherry-pick of commit 33240e87b9 from
master, but has to be done by hand because the DB format has diverged. Read
more in the commit message used there, which follows.
Due to the way we funk around with package data loading, we had a condition
where the filelist got doubled up because it was loaded twice.
Packages are originally loaded with INFRQ_BASE. In an upgrade/sync, the
package is checked for file conflicts next, leaving us in an "INFRQ_BASE |
INFRQ_FILES" state. Later, when committing a single package, we have an
explicit call to _alpm_local_db_read() with INFRQ_ALL as the level. Because
the package's level did not match this, we skipped over our previous "does
the incoming level match where I'm at" shortcut, and continued to load
things again, because of a lack of fine-grained checking for each of DESC,
FILES, and INSTALL.
The end result is we loaded the filelist twice, causing our remove logic to
iterate twice over the installed files, spewing a bunch of "cannot find file
X" messages.
Fix the problem by doing a bit more bitmasking logic throughout the load
method, and also fix the sanity check at the beginning of the function- this
should *only* be used for local packages as opposed to the "not a package"
check that was there before.
Signed-off-by: Dan McGee <dan@archlinux.org>
After all the debate as to what to do on maint, we are going to end up just
incorporating epoch into the version string, so we don't need this separate
field at all. Revert commit 5c8083baa4 and also kill the force flag we were
recording here as well.
Signed-off-by: Dan McGee <dan@archlinux.org>
Due to the way we funk around with package data loading, we had a condition
where the filelist got doubled up because it was loaded twice.
Packages are originally loaded with INFRQ_BASE. In an upgrade/sync, the
package is checked for file conflicts next, leaving us in an "INFRQ_BASE |
INFRQ_FILES" state. Later, when committing a single package, we have an
explicit call to _alpm_local_db_read() with INFRQ_ALL as the level. Because
the package's level did not match this, we skipped over our previous "does
the incoming level match where I'm at" shortcut, and continued to load
things again, because of a lack of fine-grained checking for each of DESC,
FILES, and INSTALL.
The end result is we loaded the filelist twice, causing our remove logic to
iterate twice over the installed files, spewing a bunch of "cannot find file
X" messages.
Fix the problem by doing a bit more bitmasking logic throughout the load
method, and also fix the sanity check at the beginning of the function- this
should *only* be used for local packages as opposed to the "not a package"
check that was there before.
A debug log message was added to upgraderemove as well to match the one
already in the normal remove codepath.
Signed-off-by: Dan McGee <dan@archlinux.org>
libarchive eventually calls it anyway, but backtraces make a lot more sense
if we call it, as well as matching our precedent from alpm_pkg_load().
Signed-off-by: Dan McGee <dan@archlinux.org>
This is helpful anyway to the user, and should also be helpful to us if we
see problems cropping up in the check during development.
Also add a missing ->used = 0 initialization in the code path less taken.
Signed-off-by: Dan McGee <dan@archlinux.org>
Although they won't be the same in the gettext catalog because of the '\n'
we should still use the same text.
Signed-off-by: Dan McGee <dan@archlinux.org>
This can take a while too, and it is really easy to add the necessary
callback stuff for adding a progressbar.
Signed-off-by: Dan McGee <dan@archlinux.org>
These were just two small things I came across today and found could be
fixed or helpful, so I've added them and I'm not sure what else to bundle
them with. commit_count++
Signed-off-by: Dan McGee <dan@archlinux.org>
We were checking if a package existed locally, but then using the
incoming package to calculate removed size rather than the currently
installed package.
Also adjust the local variable in the replaces loop to make it more
clear that we are always dealing with local packages here.
Signed-off-by: Dan McGee <dan@archlinux.org>
All of these can be done with integer division; the only slightly
interesting part is ensuring we round up like before with calling the
ceil() function.
We can also remove the math library from requirements; now that the only
ceil() calls are gone, we don't need this anymore.
Signed-off-by: Dan McGee <dan@archlinux.org>
file_pkg_ops can be a static struct like in other backends, we just need
to initialize it at some point.
Dan: add initialization flag.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
None of these warn at the normal "-Wall -Werror" level, but casts do occur
that we are fine with. Make them explicit to silence some warnings when
using "-Wconversion".
Signed-off-by: Dan McGee <dan@archlinux.org>
There is a lot of swtiching between size_t and int for alpm_list sizes
in the codebase. Start converting these to all be size_t by adjusting
the return type of alpm_list_count and fixing all additional warnings
given by -Wconversion that are generated by this change.
Dan: a few more small changes to ensure things compile, adjusting some
printf format string characters to accommodate the larger size on x86_64.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
All functions that are limited to the local translation unit are
declared static. This exposed that the _pkg_get_deltas declaration
in be_local.c was being satified by the function in packages.c which
when declared static caused linker failures.
Fixes all warnings with -Wmissing-{declarations,prototypes}.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
We use PATH_MAX everywhere by including limits.h so there is no
point in doing a check for it in a different header when dealing
with FreeBSD's libfetch.
Also, remove autoconf check for strings.h header as it is not used
anywhere.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
POSIX does not require PATH_MAX be defined when there is not actual
limit to its value. This affects HURD based systems. Work around
this by defining PATH_MAX to 4096 (as on Linux) when this is not
defined.
Also, clean up inclusions of limits.h and remove autoconf check for
this header as we do not use macro shields for its inclusion anyway.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Just like we did for package name comparsions, if we add a depend name_hash
field on depend struct initialization, we can use it instead of doing a
string name comparison, saving us a lot of checks in the depcmp code.
Signed-off-by: Dan McGee <dan@archlinux.org>
Noticed when tweaking testdb, when we run _alpm_depcmp in loops and call it
seven million times, the strdup()/free() combo can add up. Remove the need
for any string duplication by some pointer manipulation and use of strncmp
instead of strcmp. Also kill the function logger and add an escape so we
don't needlessly retrieve the list of provides.
Signed-off-by: Dan McGee <dan@archlinux.org>
The old function was written in a time before we relied on it for nearly
every operation. Since then, we have switched to the archive backend and now
fast parsing is a big deal.
The former function made a per-character call to the libarchive
archive_read_data() function, which resulted in some 21 million calls in a
typical "load all sync dbs" operation. If we instead do some buffering of
our own and read the blocks directly, and then find our newlines from there,
we can cut out the multiple layers of overhead and go from archive to parsed
data much quicker.
Both users of the former function are switched over to the new signature,
made easier by the macros now in place in the sync backend parsing code.
Performance: for a `pacman -Su` (no upgrades available),
_alpm_archive_fgets() goes from being 29% of the total time to 12% The time
spent on the libarchive function being called dropped from 24% to 6%.
This pushes _alpm_pkg_find back to the title of slowest low-level function.
Signed-off-by: Dan McGee <dan@archlinux.org>
The amount of diskspace needed for a transaction can be less than
zero. Only test this against the available disk space if it is
positive, which avoids a comparison being made between signed and
unsigned types (-Wsign-compare).
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
Always declare a function with (void) rather than () when we expect
no arguements. Fixes all warnings with -Wstrict-prototypes.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This simplifies a lot of the repetative code and makes it obvious where the
tricky or different ones are (e.g. depends, dates). It also makes it
significantly easier to change the way this code works in the future.
There should be no functional change with this patch.
Signed-off-by: Dan McGee <dan@archlinux.org>
Saves a few bytes due to padding (256 -> 248 bytes), especially on x86_64,
so we get the overhead of our new hash field right back.
Signed-off-by: Dan McGee <dan@archlinux.org>
This results in huge gains to a lot of our codepaths since this is the most
frequent method of random access to packages in a list. The gains are seen
in both profiling and real life.
$ pacman -Sii zvbi
real: 0.41 sec -> 0.32 sec
strcmp: 16,669,760 calls -> 473,942 calls
_alpm_pkg_find: 52.73% -> 26.31% of time
$ pacman -Su (no upgrades found)
real: 0.40 sec -> 0.50 sec
strcmp: 19,497,226 calls -> 524,097 calls
_alpm_pkg_find: 52.36% -> 26.15% of time
There is some minor risk with this patch, but most of it should be avoided
by falling back to strcmp() if we encounter a package with a '0' hash value
(which we should not via any existing code path). We also do a strcmp once
hash values match to ensure against hash collisions. The risk left is that a
package name is modified once it was originally set, but the hash value is
left alone. That would probably result in a lot of other problems anyway.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is prepping for the addition of a hash field to each package to greatly
speed up the string comparisons we frequently do on package name in
_alpm_pkg_find.
Signed-off-by: Dan McGee <dan@archlinux.org>
