1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
|
== This is my custom TODO file ==
* Dan and I both have personal TODOs. Perhaps we should kill off TODO and
TODO.autoconf
* transaction object should contain two package list (install and remove)
instead of a single list of syncpkgs - this should allow us to get rid of that
type. This also requires seperate functionality to return a list of
"replaces" packages to the front end, so the frontend can handle the QUESTION()
stuff in that case
* _alpm_splitdep accepts a pre-allocated argument, instead of returning the
actual object like all the other function do... we should make the scheme the
same: pmdepend_t *_alpm_splitdep(const char *depstring);
* Look into other VCSs to use. The main CVS repo will remain, but having a
distributed system to allow for easy patching/pushing/pulling would be nice
- monotone and mercurial look like the top contenders in my book, but I need
to evaluate both a bit more.
* src/pacman:
- There's quite a few single function headers which contain the pacman_*
functions. We should move these to a single header (pacman.h) to clean up
the source a bit.
* lib/libalpm:
- move logging (alpm_logaction) out of the library. log files should be
written by the app, not the library. Adding a PM_LOG_* constant that
frontends can recognize and send to a file is probably a good idea.
- clear up list allocation/deallocation - some lists need to be free'd, some
do not and there is no clear indication WHEN this should happen.
- remove DB entries (directories) on a read error?
- Add a pm_errstr error string, which will allow us to do things like:
pm_errno = PM_ERR_LIBARCHIVE;
pm_errstr = archive_error_string(archive);
or:
pm_errno = PM_ERR_LIBDOWNLOAD;
pm_errstr = downloadLastErrString;
This way we don't break abstraction when returning specific errors in
cases like the above.
* pacman: A LOT of functions are way too long. There should be an upper limit of
100-200 lines. _alpm_add_commit is around 600 lines, and is far too complex.
Quite a lot of this needs to be refactored as we go.
Function list (from Dan McGee):
add.c:
_alpm_add_commit
be_files.c:
_alpm_db_read
_alpm_db_write
conflict.c:
_alpm_checkconflicts
_alpm_db_find_conflicts
package.c:
_alpm_pkg_load (getting a bit long)
remove.c:
_alpm_remove_commit
server.c:
_alpm_downloadfiles_forreal
sync.c:
_alpm_sync_sysupgrade (getting there)
_alpm_sync_prepare
_alpm_sync_commit
* pacman: fixup doxygen documentation for public interface
* libalpm: just because a function is in alpm.h doesn't mean it needs to be in
alpm.c - we should move functions around where they should be. In fact,
alpm.c might not be needed at all, if things were organized properly.
* feature for 3.1: package file hooks *
I've been planning on this one for some time. Here's a simple rundown:
in /etc/pacman.d/hooks:
Hook /usr/include/* : /usr/bin/ctags -R /usr/include/*.h -f /usr/include/systags
This will allow us to make "global hooks" to simplify a lot of repetitive
install files (scrollkeeper, depmod, etc). This also allows us to move
ldconfig out of pacman entirely.
possible: /etc/pacman.hooks/* files for hooks, so packages can add them too
* feature for 3.1: multiple search/match types
option: MatchType regex (current way)
MatchType fnmatch (use fnmatch to match things like 'pacman -S gnome*')
MatchType plain (no matching. expect plain text).
* feature for 3.1: revamp the autotools system. I'd LOVE to use a manual system
like wmii and friends do. It'd be real nice if we could just do away with
autotools altogether.
**** BUGS ****
* Removal of conflicts on -A and -U (INCOMPLETE)
* pacman: add pactest cases/functionality for all -Q operations and make sure
the output is sane
|
