| 1 |
-----BEGIN PGP SIGNED MESSAGE----- |
| 2 |
Hash: SHA1 |
| 3 |
|
| 4 |
Hey there, |
| 5 |
|
| 6 |
I saw some post recently about adding more virtual for vi or wine |
| 7 |
or... his is probably not going to stop. Each time a fork exists or a |
| 8 |
equivalent program exists user will probably want the flexibility of one vs |
| 9 |
the other. |
| 10 |
|
| 11 |
The problem with virtual today is that adding one require to change |
| 12 |
all ebuilds that made use of the packages the virtual defines. This seems=20 |
| 13 |
overkill to me. While at the same time making it transparent would affect |
| 14 |
all ebuild at the same time and would require huge amount of QA. |
| 15 |
|
| 16 |
If instead of virtual we would have 'equivalent' or 'same' or 'alias' |
| 17 |
then we could add new virtual without changing the ebuild for which no change |
| 18 |
are require other than dep changes. |
| 19 |
|
| 20 |
For example, if a package was to depend on vi and an alias of vi |
| 21 |
existed for nano, the ebuild would first attempt to look for vi if not found |
| 22 |
would consult the alias db and find out that nano may as well fit for the |
| 23 |
requirement. Alias could even be made to be ebuild specific, for example, |
| 24 |
this alias for x package is valid only if you're compiling y... |
| 25 |
|
| 26 |
We could also add the functionnality to emerge (if it doesn't have |
| 27 |
already) to say that a package is not install even if it is (the reverse of |
| 28 |
inject) so to let the user control which alias is used for a specific package |
| 29 |
if he already have more than one installed. |
| 30 |
|
| 31 |
So from my example, if you had vi AND nano installed but want package |
| 32 |
y to be compiled with nano and not vi you would do emerge reject vi y.ebuild. |
| 33 |
Other interesting stuff would be that when doing emerge -vp y.ebuild you |
| 34 |
would be able to see the following |
| 35 |
|
| 36 |
[ebuild N ] app-text/vi (alias app-text/nano, ..., ...) |
| 37 |
[ebuild N ] a-b/y-1.0 |
| 38 |
|
| 39 |
The system would even be backward-compatible with today virtual by |
| 40 |
changing all virtual to aliases and the package they represent as aliases |
| 41 |
themselves... |
| 42 |
- -- |
| 43 |
|
| 44 |
Yannick Koehler |
| 45 |
Software Designer - Colubris Networks Inc. |
| 46 |
|
| 47 |
Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD34D4575 |
| 48 |
Key fingerprint = 8387 F009 A55F FFCD F87D BE18 6DEA 8D06 D34D 4575 |
| 49 |
|
| 50 |
-----BEGIN PGP SIGNATURE----- |
| 51 |
Version: GnuPG v1.2.2 (GNU/Linux) |
| 52 |
|
| 53 |
iD8DBQE/DVnDbeqNBtNNRXURAhV5AJ0QZO2ThciL90URYodvfrSgJilxbwCfbK3R |
| 54 |
FAMPvEA/JIR4uiCRiq0blvQ= |
| 55 |
=hTpH |
| 56 |
-----END PGP SIGNATURE----- |
| 57 |
|
| 58 |
-- |
| 59 |
gentoo-dev@g.o mailing list |
Archives