| index : pacman | |
| Archlinux32 fork of pacman | gitolite user |
| summaryrefslogtreecommitdiff |
| -rw-r--r-- | TODO.dan | 101 |
diff --git a/TODO.dan b/TODO.dan deleted file mode 100644 index a6c33682..00000000 --- a/TODO.dan +++ /dev/null @@ -1,101 +0,0 @@ -TODO.dan -======== - -This is my personal TODO list. No guarantees if it is out of date and things no -longer apply. If you want to help out with any of it, send me an email. - -Pacman 3.1: ------------ - -Downgrade feature - allow users to see cached packages and downgrade to -(previous or any?) available options. - -Installed size and download size in -Si/Qi output should scale with package -size- KB to MB to GB. We should also get consistancy of K/KB, M/MB, etc. - -Extreme similarity between some of the sync and add code...we have to be able -to abstract more away from sync actions and add actions to just 'actions' -(example: sync,c, add.c, and deptest.c all contain a switch on PM_DEP_MOD_*). -Merge update, add, and sync code when possible, so we reduce duplication. - -Review progress/transaction event subsystem. (from TODO) - -Add Doxygen comments to every function in libalpm including private functions. -Ensure functions are doing exactly their purpose and nothing more. Find -functions that perform duplicate behavior and merge them. Combine and refactor -others. Delete unnecessary functions. Stop keeping everything clustered in -little add/sync/remove parts, allow transactions to share code. - -Go through options list. Decide if namings are good, all options are still -relevant, etc. Ideas for -Re (#6273), changing meaning of -c (has two meanings, -another FS bug), etc. Remove the -A flag and possibly -D, -T, and -Y (-Y is -killed now in favor of vercmp binary) if they can be done by other actions. -Possible switch of -U --> -I (#5571). - -Review display and logging functions. There seems to be an abundance of them. -Make it extensible to use color, review what verbose/debug means. Perhaps -separate logging functionality- Pacman has its normal log, and alpm backend -keeps a very simple log file - listing only adds (including syncs) and removes. -This way a consistency list can be kept of what is currently installed and what -isn't, without all the logging junk from the front end. - -Profile the code. Find the functions that are being called a lot, and see what -can be done about it. Find out if all these calls are necessary (e.g. excessive -alpm_list_count calls), and maybe think about changing data structures to speed -operations up (e.g. a 'count' field). NOTE: already found two huge issues with -quick profiling- excessive setlocale calls, and extremely excessive strcmp -calls. - -Fix other localized issues- use non-printf when necessary. We may need to use -some wchar_t output on the progress bar as char/byte counts differ here. Sizes -of packages (e.g. 10,400.23 MB) should all be localized with correct -seperators. - -Rewrite makepkg to use terminal-safe coloring/bolding. tput utility should -allow us to do this. Make universal message functions for systemwide use, -including all pacman utilites- abs, pacman-optimize, etc. - -Bugs/FRs to smash: 6468, 6437, 6430?, 6420, 6404, 6389, 6312?, 6284, 6273?, -6255?, 6208, 5987, 5885, 5571, 4182, 3492, 2810?, 1769, 1588, 1571 - -Update doxygen comments, they may need some work. Try to document all of the -private internal functions too- it helps a ton for people just getting a -start on pacman hacking. - -Other smaller things: ---------------------- - -unsigned int vs. unsigned- determine a standard and stick with it - -FS #4185 - move where message is logged, perhaps? - -Update copyrights (2007) - -Update pacman website, and add/finish pacman coding style page - -Refine makepkg error codes. Each kind of failure could have its own code: - --package already built - --failed integ checks - --failed build - --etc. - -Add utility function to either frontend or backend to convert sizes: e.g. bytes -to KB, MB, GB. - -Revamp the downloadprog function a bit. Seems kind of messy. - ---print-uri option to sync should not require saying yes or no to up to date - -Build a replacement for this, or at least standardize its use. We shouldn't -always need to pass handle->root around, it is constant. Something like char* -buildpath(file). - /* build the new entryname relative to handle->root */ - snprintf(filename, PATH_MAX, "%s%s", handle->root, entryname); - -Random ideas: -------------- - -Possibly split utilities/extras from pacman package - -Make .indent.pro file for GNU indent (started, but didn't finish) - |