Only load filesystem details for the mount points that we're actually
going to write to. This reduces our syscall count considerably. In the
case of installation, we would actually stat every mountpoint twice (an
extra round for download diskspace) which means (on my system) a total
of 60 syscalls to write to 3 partitions when installing the kernel
package. This change reduces the 60 syscalls down to the expected 3.
A slight debug output change is added here to discern between a
mountpoint added to our linked list versus when we actually load the fs
info.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
add mount_point_load_fsinfo() for platforms using getmntent().
Dan: move the #ifdef slightly so we don't have unused functions on
certain platforms (e.g., OS X).
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
With lazy loading in place, it's now quite obvious that we aren't
necessarily checking the right mountpoint for necessary download space.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Not sure why this one wasn't showing up on x86_64, but this fixes the
compile on i686.
diskspace.c: In function 'calculate_removed_size':
diskspace.c:247:4: error: assuming signed overflow does not occur when negating a division [-Werror=strict-overflow]
cc1: all warnings being treated as errors
Signed-off-by: Dan McGee <dan@archlinux.org>
If one had a mountpoint at '/e' (don't ask), a file being installed to
'/etc' would map to it incorrectly. Ensure we do more than just prefix
matching on paths by doing some more sanity checks once the simple
strncmp() call succeeds.
Signed-off-by: Dan McGee <dan@archlinux.org>
Was able to get my hands on one of these boxes today, so add yet another
new way of doing this. I'm glad these calls are so standardized. This
was compile tested on Linux and Illumos and seems to still be working in
both places.
Signed-off-by: Dan McGee <dan@archlinux.org>
Ensures that config.h is always ordered correctly (first) in the
includes. Also means that new source files get this for free without
having to remember to add it.
We opt for -imacros over -include as its more portable, and the
added constraint by -imacros doesn't bother us for config.h.
This also touches the HACKING file to remove the explicit mention of
config.h as part of the includes.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This logic is reused in both diskspace and downloadspace check
functions, so pull it out into its own static method.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This function determines if the given cachedir has at least the given
amount of free space on it. This will be later used in the sync code to
preemptively halt downloads.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
This will be useful when extending disk space checks to free space
checking before we download package files.
Signed-off-by: Dan McGee <dan@archlinux.org>
This adjusts type usage to match POSIX provided types from
<sys/types.h> rather than assuming everything will fit in a long or
unsigned long. Use fsblkcnt_t (unsigned) and blkcnt_t (signed) as
appropriate. These are affected the same way off_t is on 32 bit
platforms, where the types are extende to 64 bits if large file support
is enabled.
Because most numbers here are block counts, this isn't
near as pressing as using a 32-bit variable for file sizes where
anything over 2GiB can burn you; we likely can support files at least
512 but mainly 4096 times larger.
Signed-off-by: Dan McGee <dan@archlinux.org>
Put all the callback stuff in alpm.h in one spot, and make the following
renames for clarity with the new structure:
ALPM_TRANS_EVT_* --> ALPM_EVENT_*
ALPM_TRANS_CONV_* --> ALPM_QUESTION_*
ALPM_TRANS_PROGRESS_* --> ALPM_PROGRESS_*
alpm_option_get_convcb() --> alpm_option_get_questioncb()
alpm_option_set_convcb() --> alpm_option_set_questioncb()
Signed-off-by: Dan McGee <dan@archlinux.org>
This was just disgusting before, unnecessary to limit these to only
usage in a transaction. Still a lot of more room for cleanup but we'll
start by attaching them to the handle rather than the transaction we may
or may not even want to use these callbacks.
Signed-off-by: Dan McGee <dan@archlinux.org>
It would prevent compilation of pacman on FreeBSD, and possibly other
systems.
Signed-off-by: Rémy Oudompheng <remy@archlinux.org>
Signed-off-by: Dan McGee <dan@archlinux.org>
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>
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>
Only one of these looked like a real red flag, in find_requiredby(), but
it doesn't hurt to fix several of them up anyway.
Unfortunately, we can't turn this on universally due to things like the
sync(), remove(), etc. builtins which we often use as variable names.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is the last user of our global handle object. Once again the diff
is large but the functional changes are not.
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>
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>
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>
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>
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>
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>
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>
Allan McRae