| 1 |
Jason Stubbs posted <200509060839.34718.jstubbs@g.o>, excerpted |
| 2 |
below, on Tue, 06 Sep 2005 08:39:32 +0900: |
| 3 |
|
| 4 |
> On Tuesday 06 September 2005 01:06, Philip Webb wrote: |
| 5 |
>> 050905 Jason Stubbs wrote: |
| 6 |
>> > it's possible that unmerging slotted packages of the one key |
| 7 |
>> > may break your system. How's about not warning |
| 8 |
>> > if there's more than one installed cat/pkg (rather than cat/pkg-ver) |
| 9 |
>> > satisfying the profile atom that is being triggered? |
| 10 |
>> |
| 11 |
>> -- patch snipped -- |
| 12 |
>> |
| 13 |
>> I'ld say that the behaviour should be left alone |
| 14 |
>> pending a larger rewrite of Portage's handling of this kind of thing. |
| 15 |
>> Meanwhile, simply amend the warning to read in red letters eg |
| 16 |
> |
| 17 |
> Why? |
| 18 |
> |
| 19 |
>> "WARNING : removing this package may break your system !! |
| 20 |
>> Have you checked that you have a proper alternative installed ?? |
| 21 |
>> If you are not certain what you are doing, please STOP NOW !! |
| 22 |
>> |
| 23 |
>> Do you want to proceed (NO/yes) ? " |
| 24 |
> |
| 25 |
> No. |
| 26 |
|
| 27 |
Expanding slightly, since no reasons were given. |
| 28 |
|
| 29 |
Portage is normally non-interactive. The only way the above would work |
| 30 |
would be if it were somehow keyed into the "ask" parameter, and then, it |
| 31 |
would need to be run before any dependencies were handled, a rather less |
| 32 |
than workable possibility at this point (when merging anyway, tho this |
| 33 |
particular thing is unmerging). |
| 34 |
|
| 35 |
Currently, the only way to handle this sort of thing is with a |
| 36 |
time-delay/beep warning, or by up and dying, in /extreme/ cases, with a |
| 37 |
message like *THIS WILL BREAK YOUR SYSTEM UNLESS YOU DO THIS AND THIS |
| 38 |
FIRST!* Do that, and set environmental variable |
| 39 |
WONT_BREAK_SYSTEM_NOW=DOIT, to merge the package. |
| 40 |
|
| 41 |
That only works AFTER dependencies may have been merged, so it's not |
| 42 |
optimal, either, but in a very few extreme cases, that's the sort of |
| 43 |
solution that has been used. |
| 44 |
|
| 45 |
Supposedly, portage is to be a bit better at handling this sort of thing, |
| 46 |
in a future version, when it will be able to do it before merging |
| 47 |
dependencies. However, until then, choices are somewhat limited. IMO, a |
| 48 |
warning about system profile stuff SHOULD sound appropriately drastic. |
| 49 |
Once the admin is confident that it's doable without issue, then they can |
| 50 |
go ahead, and if their confidence was misplaced, then it's their issue. |
| 51 |
The only remaining thing then, is to put enough info into the warning to |
| 52 |
let the admin take the appropriate action. Pointing out the virtual it's |
| 53 |
affecting seems like enough info to me. If that's not enough for some |
| 54 |
folks, perhaps they'd be better off keeping it around, just in case. |
| 55 |
|
| 56 |
-- |
| 57 |
Duncan - List replies preferred. No HTML msgs. |
| 58 |
"Every nonfree program has a lord, a master -- |
| 59 |
and if you use the program, he is your master." Richard Stallman in |
| 60 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
| 61 |
|
| 62 |
|
| 63 |
-- |
| 64 |
gentoo-dev@g.o mailing list |
Archives