This accomplishes quite a few things with one rather invasive change.
1. Iteration is much more performant, due to a reduction in pointer
chasing and linear item access.
2. Data structures are smaller- we no longer have the overhead of the
linked list as the file struts are now laid out consecutively in
memory.
3. Memory allocation has been massively reworked. Before, we would
allocate three different pieces of memory per file item- the list
struct, the file struct, and the copied filename. What this resulted
in was massive fragmentation of memory when loading filelists since
the memory allocator had to leave holes all over the place. The new
situation here now removes the need for any list item allocation;
allocates the file structs in contiguous memory (and reallocs as
necessary), leaving only the strings as individually allocated. Tests
using valgrind (massif) show some pretty significant memory
reductions on the worst case `pacman -Ql > /dev/null` (366387 files
on my machine):
Before:
Peak heap: 54,416,024 B
Useful heap: 36,840,692 B
Extra heap: 17,575,332 B
After:
Peak heap: 38,004,352 B
Useful heap: 28,101,347 B
Extra heap: 9,903,005 B
Several small helper methods have been introduced, including a list to
array conversion helper as well as a filelist merge sort that works
directly on arrays.
Signed-off-by: Dan McGee <dan@archlinux.org>
As noted by Allan, we failed pretty hard if gpgme was compiled out. With
these changes, only sign001.py fails. This can/will be fixed later once
we beef up the test suite with more signing tests anyway.
Signed-off-by: Dan McGee <dan@archlinux.org>
If we can't read the keyring, gpgme will output confusing debug
information and fail to verify the signature, so we should log some
debug information.
Signed-off-by: Florian Pritz <bluewind@xinu.at>
Signed-off-by: Dan McGee <dan@archlinux.org>
This is a wrapper function for access() which logs some debug
information and eases handling in case of split directory and filename.
Signed-off-by: Florian Pritz <bluewind@xinu.at>
Signed-off-by: Dan McGee <dan@archlinux.org>
This addresses FS#25141. We shouldn't remove every empty directory we
come across during the removal process unless it is truly not known to
any other package. This will prevent removal of essential directories
such as '/var/lock/'.
This is accomplished by first checking the empty/non-empty status of a
directory, which was previously done implicitly by calling rmdir() and
ignoring errors. We do this to avoid the next (new) check in most cases,
which is to look at all local packages to see if the to-be-removed
directory is present in another packages' filelist. If we do not find it
anywhere, then we remove it, else we keep the file around.
The pactest has been updated to test more cases, as well as finding a
flaw in the original expected to fail case- we need separate DIR and
FILE based EXIST rules.
Signed-off-by: Dan McGee <dan@archlinux.org>
This can only ever operate on the local database, and a local package at
that. Change the function signature to take a handle and package object,
add the relevant asserts, and ensure the frontend can detect the package
not found condition when finding packages to pass to this method.
Signed-off-by: Dan McGee <dan@archlinux.org>
The bulk of this commit is adding new tests to ensure the new behavior
works without disrupting old behavior. This is a relatively sane maneuver
when a package adds a conf file (e.g. '/etc/mercurial/hgrc') that was
not previously in the package, but it is placed in the backup array. In
essence, we can treat the existing file as having always been a part of
the package and do our normal compare/install as pacnew logic checks.
Signed-off-by: Dan McGee <dan@archlinux.org>
This code duplication has always been a rather clumsy casuality of
fixing some past upgrade issues. Unify the removal code across upgrade
and remove operations into a new _alpm_remove_single_package() method
wihch makes it very clear how we handle upgrade and remove differently,
via several conditionals on newpkg.
This commit highlights interesting behavior such as the fact that the
implicit removal in every package upgrade never gets transaction events
or progress callbacks.
Signed-off-by: Dan McGee <dan@archlinux.org>
Fixes "error: no previous prototype for '_alpm_raw_cmp'
[-Werror=missing-prototypes]" warnings, and also prevents someone from
getting the prototypes and functions out of sync.
Signed-off-by: Dan McGee <dan@archlinux.org>
Restore some sanity to the number of arguments passed to _alpm_download
and curl_download_internal.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
This means creating a new struct which can pass more descriptive data
from the back end sync functions to the downloader. In particular, we're
interested in the download size read from the sync DB. When the remote
server reports a size larger than this (via a content-length header),
abort the transfer.
In cases where the size is unknown, we set a hard upper limit of:
* 25MiB for a sync DB
* 16KiB for a signature
For reference, 25MiB is more than twice the size of all of the current
binary repos (with files) combined, and 16KiB is a truly gargantuan
signature.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
URLs might end with a slash and follow redirects, or could be a
generated by a script such as /getpkg.php?id=12345. In both cases, we
may have a better filename that we can write to, taken from either
content-disposition header, or the effective URL.
Specific to the first case, we write to a temporary file of the format
'alpmtmp.XXXXXX', where XXXXXX is randomized by mkstemp(3). Since this
is a randomly generated file, we cannot support resuming and the file is
unlinked in the event of an interrupt.
We also run into the possibility of changing out the filename from under
alpm on a -U operation, so callers of _alpm_download can optionally pass
a pointer to a *char to be filled in by curl_download_internal with the
actual filename we wrote to. Any sync operation will pass a NULL pointer
here, as we rely on specific names for packages from a mirror.
Fixes FS#22645.
Signed-off-by: Dave Reisner <d@falconindy.com>
The supposed safety blanket of this function is better handled by
explicit length checking and usages of strlen() on known NULL-terminated
strings rather than hoping things fit in a buffer. We also have no need
to fully fill a PATH_MAX length variable with NULLs every time as long
as a single terminating byte is there. Remove usages of it by using
strcpy() or memcpy() as appropriate, after doing length checks via
strlen().
Signed-off-by: Dan McGee <dan@archlinux.org>
We can readily detect the first node in a list by checking if
node->prev->next is NULL. So there is no need to pass the head
of the list to this function and its prototype now looks like
all the other item accessors.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
The only thing this accessor did was remove the const qualifier
given our entire list implementation requires passing around the
head anyway.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
They are placeholders, but important for things like trying to re-sync a
database missing a signature. By using the alpm_db_validity() method at
the right time, a client can take the appropriate action with these
invalid databases as necessary.
In pacman's case, we disallow just about anything that involves looking
at a sync database outside of an '-Sy' operation (although we do check
the validity immediately after). A few operations are still permitted-
'-Q' ops that don't touch sync databases as well as '-R'.
Signed-off-by: Dan McGee <dan@archlinux.org>
Show output in -Qip for each package signature, which includes the UID
string from the key ("Joe User <joe@example.com>") and the validity of
said key. Example output:
Signatures : Valid signature from "Dan McGee <dpmcgee@gmail.com>"
Unknown signature from "<Key Unknown>"
Invalid signature from "Dan McGee <dpmcgee@gmail.com>"
Also add a backend alpm_sigresult_cleanup() function since memory
allocation took place on this object, and we need some way of freeing
it.
Signed-off-by: Dan McGee <dan@archlinux.org>
The error code is in fact a bitmask value of an error code and an error
source, so use the proper function to get only the relevant bits. For
the no error case, this shouldn't ever matter, but it bit me when I was
trying to compare the error code to other values and wondered why it
wasn't working, so set a good example.
Signed-off-by: Dan McGee <dan@archlinux.org>
This gives us more granularity than the former Never/Optional/Always
trifecta. The frontend still uses these values temporarily but that will
be changed in a future patch.
* Use 'siglevel' consistenly in method names, 'level' as variable name
* The level becomes an enum bitmask value for flexibility
* Signature check methods now return a array of status codes rather than
a simple integer success/failure value. This allows callers to
determine whether things such as an unknown signature are valid.
* Specific signature error codes mostly disappear in favor of the above
returned status code; pm_errno is now set only to PKG_INVALID_SIG or
DB_INVALID_SIG as appropriate.
Signed-off-by: Dan McGee <dan@archlinux.org>
Now that the filelists capture mode and size information, we can read
the data from there and prevent having to loop through and uncompress
every archive to check required diskspace usage.
Signed-off-by: Dan McGee <dan@archlinux.org>
This allows us to capture size and mode data when building filelists
from package files. Future patches will take advantage of this newly
available information, and frontends can use it as well.
Signed-off-by: Dan McGee <dan@archlinux.org>
This saves replicating the potentially large list of files in a package
that is being removed.
Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
We passed in 'line', but not 'buf.line'. In addition, the macros
building off of READ_NEXT() assume variable names anyway. Since we only
use these macros in one function, might as well simplify them.
Signed-off-by: Dan McGee <dan@archlinux.org>
Change the check into a loop over all signatures present and returned by
GPGME. Also modify the return values and checks slightly now that I know
a little bit more about what type of values are returned.
Signed-off-by: Dan McGee <dan@archlinux.org>
There is little need to expose the guts of this function even within the
library. Make it static in be_local.c, and clean up a few other things
since we know exactly where it is being called from:
* Remove unnecessary origin checks in _cache_get_*() methods- if you are
calling a cache method your package type will be correct.
* Remove sanity checks within local_db_read() itself- packages will
always have a name and version if they get this far, and the package
object will never be NULL either.
The one case calling this from outside the backend was in add.c, where
we forced a full load of a package before we duplicated it. Move this
concern elsewhere and have pkg_dup() always force a full package load
via a new force_load() function on the operations callback struct.
Signed-off-by: Dan McGee <dan@archlinux.org>
Some of these are legit (the backup hash NULL checks), while others are
either extemely unlikely or just impossible for the static code
analysis to prove, but are worth adding anyway because they have little
overhead.
Signed-off-by: Dan McGee <dan@archlinux.org>
