| 1 |
On 10/21/2015 09:13 AM, Rob Wortman wrote: |
| 2 |
> On 2015-10-20 at 22:11:24 -0700, zmedico@g.o wrote: |
| 3 |
>> Any packages that are not reachable from @world are ripe for removal by |
| 4 |
>> --depclean, so we allow their dependencies to break in order to satisfy |
| 5 |
>> other dependencies (like in bug 563482). If you don't use --deep, then |
| 6 |
>> emerge may try to build something that depends on one of these |
| 7 |
>> unreachable packages with broken dependencies, such that whatever you |
| 8 |
>> are trying to build has broken indirect dependencies (which is likely to |
| 9 |
>> trigger a build failure like in bug 563482). |
| 10 |
> |
| 11 |
> I think I understand. So, one could get the hypothetical scenario: |
| 12 |
> |
| 13 |
> # emerge --oneshot A # which depends on B |
| 14 |
> # emerge --update @world # shuffles stuff around breaking B |
| 15 |
> # emerge C # which depends on A |
| 16 |
> |
| 17 |
> Package C's dependency is filled, so emerge goes ahead and builds C. |
| 18 |
> Now, either C fails to build, or it is installed but fails at runtime, |
| 19 |
> because it depends on a package which depends on a package which is |
| 20 |
> broken. |
| 21 |
> |
| 22 |
> Sound about right? |
| 23 |
> |
| 24 |
|
| 25 |
Yeah, and if you run emerge --depclean regularly, then it will prevent |
| 26 |
problems like these. |
| 27 |
-- |
| 28 |
Thanks, |
| 29 |
Zac |
Archives