The patch adds 150 lines approximately. It is implemented by modifying
xbps_find_pkg_in_dict_by_name() and xbps_find_pkg_in_dict_by_pattern().
When no dictionary is returned if tries to find a virtual package by
looking at the "provides" array object, if found.
This also makes xbps_unpack_binary_pkg follow this convention by avoiding
static variables.
--HG--
branch : progress_callback
rename : bin/xbps-repo/util.c => bin/xbps-bin/util.c
So now its prototype is the following:
int xbps_humanize_number(char *buf, int64_t bytes)
It is a wrapper around NetBSD's humanize_number(3) which uses 6 digits for max
length, HN_AUTOSCALE and HN_NOSPACE|HN_DECIMAL. All users have been updated.
This function replaces xbps_repository_get_path_from_pkg_dict() and
xbps_get_binpkg_local_path(). It takes a pkg dictionary as returned
by a repository pkg index or a transaction dictionary and returns
a string with the full path to the binary pkg, either in local
repos, cachedir or remote repos.
Update all code to use this function... sorry I broke ABI compatiblity.
prop_dictionary_t xbps_repository_pool_find_pkg(const char *pkg, bool bypattern, bool best)
This function iterates over the repository pool and returns a pkg
dictionary matching its pkgname if "bypattern" is false, otherwise
if "pkg" matches the pkgdep pattern. If "best" is true, it will look
in all repos and will return the newer version, otherwise the first one
that matches the arguments specified.
With this addition supporting many small registered repos is cheap, because
it checks by-package rather than by-package-list. This also helped to removed
some duplicated code and fix some rare bugs.
Bump XBPS_RELVER to 20110118.
Rather than using a SIMPLEQ to copy the package dictionary and later on
adding all items into the sorted array, just add the pkg dictionary
into the sorted array directly and removing it from the unsorted array.
This reduces memory usage and greatly sorts all packages that aren't
installed automatically in the correct order in the transaction.
Also added a large comment explaining how the algorithm works and
what it is supposed to do and when.
