On Dec 7, 2007 10:47 AM, Allan McRae <allan.mcrae@qimr.edu.au> wrote:
Hi all,
Below is a patch that tries to improve the handling of changelog dumping. Now changelogs should also get dumped when using "pacman -Qcp <file>" syntax as well as "pacman -Qc <pkg>" (See http://bugs.archlinux.org/task/7321)
This is my first patch to the pacman/libalpm code base so it probably needs lots of checking. I stared at the code for a couple of hours before I started but I am probably using some libalpm functions wrongly. I would pay particular attention to the entirely new "alpm_pkg_get_changelog" function in libalpm/package.c.
The code builds cleanly but I can't test much more than that at the moment due to my testing box having what I think is RAM issues leaving me only my work laptop...
Thanks a lot. I will look this over tonight, but I don't see anything that would cause drastic breakage. I do wonder why we never had a pkg_get_changelog() /me shrugs
Thanks for the patch, it's good for the motivation when new people come on board.
I also say thank you for this patch, but my vote: not apply (yet). Imho I told everything in my last mail: -my main reason: this patch doesn't follow our earlier "conventions" at all (no problem, you are new here, and this can be fixed) -I also silently referred to the following fact: If I didn't misunderstand add.c "monster", now .CHANGELOG is extracted as /var/lib/.../changelog in extract_single_file(). However, extract_single_file is called in case of !(trans->flags & PM_TRANS_FLAG_DBONLY) only. So this just breaks the borderline between "package install" and "record to database". This is an other reason why pmpkg_t solution is clearer. But I would like to say a contra for my preferred implementation, too: -higher memory usage [but, if you measure the memory usage with -Qo, this is not notable at all ;-] Bye ---------------------------------------------------- SZTE Egyetemi Könyvtár - http://www.bibl.u-szeged.hu This mail sent through IMP: http://horde.org/imp/