1 |
> Maybe you should search the forums for "cruft". I remember Ed (?) Catmur once |
2 |
> posted a script or something there. |
3 |
I use a script named findcruft regularly (I think it is an improved |
4 |
version of Ed Catmur's product; I'll check when I get home). It works |
5 |
by finding all files on the filesystem (avoiding areas such as /proc, |
6 |
/sys/ /dev/, /home, and such) and comparing that to the files managed |
7 |
by Portage. The list of orphaned files is full of false positives, |
8 |
caused by, for example, packages that create files outside the |
9 |
knowledge of portage (I think many eselect modules do this for |
10 |
example). It tries to diminish the "false positives" problem by having |
11 |
a database of packages and files created by using said packages. |
12 |
|
13 |
This database of course cannot be complete, and unfortunately it seems |
14 |
to be very outdated. I would love to see more maintenance of this |
15 |
script and, more importantly, the database. |
16 |
|
17 |
In the current state, the list of orphaned files is huge, with a good |
18 |
part of it probably being false positives. Since I don't want to move |
19 |
a file without being sure it is cruft, I can only read that list and |
20 |
delete the few files that I know about and am sure are cruft. I also |
21 |
run the script regularly and save the results; each time I run the |
22 |
script I compare the list with the previous one. I normally run the |
23 |
script after I unmerge some packages. |
24 |
|
25 |
In the end, if you want a clean system it seems you either have to |
26 |
reformat every couple of years, or you have to be a walking |
27 |
encyclopedia knowing the origin of every file in the huge list that |
28 |
findcruft outputs so you know if they are cruft or not. |
29 |
|
30 |
Sad. In my opinion, this problem needs to be addressed. |
31 |
|
32 |
-- |
33 |
Software is like sex: it is better when it is free - Linus Torvalds |