Re: Package Installation takes a very long time - normal?

[Ludovic Courtès]

Leo Famulari <address@hidden> skribis:

> On Mon, Jul 10, 2017 at 09:25:27PM +0100, Niall Dooley wrote:
>> OK, so it has finished but unsuccessfully. The last lines of the output were
>> as follows but none of the packages were installed. To reach this point it
>> took about 50 mins (it started around 19:55 and finished at 20:43).
>
> That sounds annoying :(

Indeed…

>> The following packages will be installed:
>>    r-caret 6.0-76 /gnu/store/dybi25l6wr14y3cxw6xazans9vlqjlji-r-caret-6.0-76
>>    r-ggplot2 2.2.1 
>> /gnu/store/11jzzbgpg67v3yysiv6pyhi27f75dyh1-r-ggplot2-2.2.1
>>    r-rpart 4.1-11 /gnu/store/rrsr0q4q4n2wjfg26kyg4klk708f675f-r-rpart-4.1-11
>>    r-e1071 1.6-8 /gnu/store/1r4if4p10rnyj6q073yjir99mpriv1d0-r-e1071-1.6-8
>>    r-randomforest 4.6-12
>> /gnu/store/nkqqls9piqzp7pdgfrka1rz005hxj19c-r-randomforest-4.6-12
>>    r-rocr 1.0-7 /gnu/store/ldn5y4qcz6kxgqkhld69w00lvbiz9682-r-rocr-1.0-7
>>    r-class 7.3-14 /gnu/store/bqx4c9r3pzmf12vlfy2hdw8hamk8z45n-r-class-7.3-14
>> 
>> guix package: error: profile contains conflicting entries for r-mass:out
>> guix package: error:   first entry: address@hidden:out
>> /gnu/store/z0nxbji8yrkdl7n6v3wdsd8iqcj39vyk-r-mass-7.3-47
>> guix package: error:    ... propagated from address@hidden
>> guix package: error:    ... propagated from address@hidden
>> guix package: error:   second entry: address@hidden:out
>> /gnu/store/gnv11ncv2h9ij48jwy86wai3xsm7c9s5-r-mass-7.3-47
>> guix package: error:    ... propagated from address@hidden
>
> I recommend updating all the packages that are trying to propagate
> r-mass. That is, R, r-car, and r-caret. We recently made a change to
> throw an error here:
>
> https://bugs.gnu.org/27271
>
> I don't fully understand the impact of the change so I asked Ludovic to
> chime in.

Previously, Guix would have proceeded with the installation; you’d get
many warning lines just before it completes, but that’s it.

The problem is that it’s a situation where failing is safer: if we let
it proceed, the profile would have to arbitrarily choose between the
files coming from one ‘r-mass’ variant and those coming from the other
one.

The fix is to arrange to have only one ‘r-mass’ in your profile.  Here
the message tells you that one of them is propagated by ‘r’ and the
second one by ‘r-caret’, so you have to get these two in sync,
presumably by adding ‘r’ to your “guix package -i” command line (I’m not
familiar with R, maybe Ricardo can comment on this.)

Does that make sense?

To avoid these problems, you might also want to use a manifest with
“guix package -m”.

I hope this clarifies things.

Thanks for your feedback,
Ludo’.