v3 of the series, reworked WRT v2:
* Change longest_ancestor_length() to take a string_list for its
prefixes argument (instead of a PATH_SEP-separated string).
* Move the responsibility for canonicalizing prefixes from
longest_ancestor_length() to its caller in
setup_git_directory_gently_1(). This keeps
longest_ancestor_length() testable, though the test inputs now have
to be canonicalized.
* Remove function string_list_longest_prefix(), which was mainly
intended to be used in the implementation of
longest_ancestor_length() but is now unneeded.
Thanks to Junio for his comments.
I would like to explicitly point out a possible objection to this
whole enterprise. GIT_CEILING_DIRECTORIES is used to prevent git from
mucking around in certain directories, to avoid expensive accesses
(for example of remote directories). The original motivation for the
feature was a user whose home directory /home/$USER was automounted.
When git was invoked outside of a git tree, it would crawl up the
filesystem tree, eventually reaching /home, and then try to access the
paths
/home
/home/.git
/home/.git/objects
/home/objects
The last three accesses would be very expensive because the system
would attempt to automount the entries listed.
This patch series has the side effect that all of the directories
listed in GIT_CEILING_DIRECTORIES are accessed *unconditionally* to
resolve any symlinks that are present in their paths. It is
admittedly odd that a feature intended to avoid accessing expensive
directories would now *intentionally* access directories near the
expensive ones. In the above scenario this shouldn't be a problem,
because /home would be the directory listed in
GIT_CEILING_DIRECTORIES, and accessing /home itself shouldn't be
expensive. But there might be other scenarios for which this patch
series causes a performance regression.
Another point: After this change, it would be trivially possible to
support non-absolute paths in GIT_CEILING_DIRECTORIES; just remove the
call to is_absolute_path() from normalize_ceiling_entry(). This might
be convenient for the test suite.
Michael Haggerty (8):
Introduce new static function real_path_internal()
real_path_internal(): add comment explaining use of cwd
Introduce new function real_path_if_valid()
longest_ancestor_length(): use string_list_split()
longest_ancestor_length(): take a string_list argument for prefixes
longest_ancestor_length(): require prefix list entries to be
normalized
normalize_ceiling_entry(): resolve symlinks
string_list_longest_prefix(): remove function
Documentation/technical/api-string-list.txt | 8 ---
abspath.c | 105 ++++++++++++++++++++++------
cache.h | 3 +-
path.c | 46 ++++++------
setup.c | 32 ++++++++-
string-list.c | 20 ------
string-list.h | 8 ---
t/t0060-path-utils.sh | 41 ++++-------
t/t0063-string-list.sh | 30 --------
test-path-utils.c | 8 ++-
test-string-list.c | 20 ------
11 files changed, 157 insertions(+), 164 deletions(-)
--
1.7.11.3
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html