| 1 |
On 3/11/20 9:43 PM, Matt Turner wrote: |
| 2 |
> On Wed, Mar 11, 2020 at 9:36 PM Matt Turner <mattst88@g.o> wrote: |
| 3 |
>> |
| 4 |
>> On Wed, Mar 11, 2020 at 9:31 PM Zac Medico <zmedico@g.o> wrote: |
| 5 |
>>> We can't can't continue above, since that will skip all of the filters |
| 6 |
>>> that occur later in the loop. So, we have to nest the below changed-deps |
| 7 |
>>> code under if options['changed-deps']: |
| 8 |
>> |
| 9 |
>> I'm happy to make that change, but I don't think it's necessary, |
| 10 |
>> strictly speaking, since this is inside an 'if not destructive' |
| 11 |
>> conditional and the only filter afterwards is 'if destructive'. |
| 12 |
> |
| 13 |
> Wait... the logic was if not destructive and |
| 14 |
> package-exists-in-porttree -> continue and do not add it to the dead |
| 15 |
> package list. |
| 16 |
> |
| 17 |
> I've just changed it so it does that if changed-deps is not set... so |
| 18 |
> keep the current behavior without --changed-deps. |
| 19 |
> |
| 20 |
> And if --changed-deps, check the porttree vs binpkg dependencies, and |
| 21 |
> if they still match, skip. |
| 22 |
|
| 23 |
Yeah, you're right. |
| 24 |
|
| 25 |
> What is wrong with that logic? |
| 26 |
|
| 27 |
The coupling with --destructive logic complicates matters. It raises the |
| 28 |
question, why isn't --time-limit logic also coupled with --destructive |
| 29 |
logic? I think "in order to preserve the status quo" is a reasonable |
| 30 |
answer to this question. |
| 31 |
-- |
| 32 |
Thanks, |
| 33 |
Zac |
Archives