On 27/04/2015 00:23, Michael Orlitzky wrote:
> On 04/26/2015 05:28 PM, Philip Webb wrote:
>>
>>> More seriously, once you start working on (3), you'll realize
>>> that just because the error msgs suck doesn't mean you can make them better.
>>
>> If they "suck", they're not worth issuing, are they ?
>> I'm not willing to become a dev, so I'll never know if I cd improve them,
>> but it doesn't follow that no-one else could.
>>
> 
> They give you the information you need to update your question.
> 
> You don't need to be a Gentoo developer to contribute to portage. They
> have a mailing list and all patches are posted there for review, Gentoo
> dev or not.
> 
> 
>>> If you're willing to wait an hour, it might be able to come up
>>> with a list of ways you could resolve a conflict, but basically
>>> all of them will be wrong, eg suggestion #1, uninstall everything.
>>
>> Really, this is a flippant response to a serious issue,
>> which is being raised more often on the Gentoo User list.
>>
> 
> It wasn't meant that way -- I was trying to point out that this is one
> of those problems that sounds easy but turns out to be incredibly hard.
> 
> Dependency resolution is already slow when it only takes your installed
> packages into account. It would take oh-so-much longer if you wanted to
> consider "what if" questions involving the entire tree. And most of the
> suggestions it would come up with are indeed ridiculous. Uninstalling a
> few things in @world will probably solve your conflict. Is that not a
> valid suggestion in some cases? Why not? Can you determine those cases
> automatically without input from the user?
> 
> I'm not saying it can't be done, but it's deceptively hard to
> (automatically) come up with a list of non-ridiculous suggestions before
> the user in question dies of old age. Relevant xkcd:
> 
>   https://xkcd.com/1425/
> 
> 


I'm aware of the scope of the problem, and I'm not asking portage to
infer what I might want or suggest solutions I didn't ask for. Besides,
"what I want" is already unambiguously defined by world and the contents
of /etc/portage/.

I'd be much happier if portage took the information *it already has* and
formatted it's output as something a tad more parseable to human brains.
Right now what it's doing is the equivalent of a core dump with an
attitude of "ah, fuck it, I give up. Here, you figure this shit out."


-- 
Alan McKinnon
alan.mckin...@gmail.com


Reply via email to