2 Commits

Author	SHA1	Message	Date
NIIBE Yutaka	29c7fb4053	gpg: Fix getting User ID. ... * g10/getkey.c (user_id_db): Remove, as no use anymore. (get_user_id_string): Use cache_get_uid_bykid. (get_user_id_byfpr): Use cache_get_uid_byfpr. * g10/objcache.c (cache_get_uid_byfpr): New. * g10/objcache.h (cache_get_uid_byfpr): New. Fixes-commit: 64a5fd3727 Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>	2019-07-11 12:32:44 +09:00
Werner Koch	64a5fd3727	gpg: New caching functions. ... * g10/objcache.c: New. * g10/objcache.h: New. * g10/Makefile.am (common_source): Add them. * g10/gpg.c: Include objcache.h. (g10_exit): Call objcache_dump_stats. * g10/getkey.c: Include objcache.h. (get_primary_uid, release_keyid_list): Remove. (cache_user_id): Remove. (finish_lookup): Call the new cache_put_keyblock instead of cache_user_id. (get_user_id_string): Remove code for mode 2. (get_user_id): Implement using cache_get_uid_bykid. -- This generic caching module is better than the ad-hoc code we used in getkey.c. More cleanup in getkey is still required but it is a start. There is also a small performance increase with the new cache: With a large keyring and --list-sigs I get these numbers: | | before | after | |------+------------+------------| | real | 14m1.028s | 12m16.186s | | user | 2m18.484s | 1m36.040s | | sys | 11m42.420s | 10m40.044s | Note the speedup in the user time which is due to the improved cache algorithm. This is obvious, because the old cache was just a long linked list; the new cache are two hash tables. Signed-off-by: Werner Koch <wk@gnupg.org>	2019-04-13 11:48:58 +02:00