bug-guix
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

bug#32749: package-with-explicit-inputs leaks-in additional inputs


From: Ludovic Courtès
Subject: bug#32749: package-with-explicit-inputs leaks-in additional inputs
Date: Tue, 18 Sep 2018 16:38:01 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)

Hello,

Jan Nieuwenhuizen <address@hidden> skribis:

> Ludovic Courtès writes:
>
>> The difference comes from the fact that ‘gnu-make-explicit-inputs’ has
>> Guile in its ‘inputs’:
>
> Ah, I missed that!
>
>> scheme@(gnu packages pawei)> (package-direct-inputs gnu-make-explicit-inputs)
>> $5 = (("libc" #<package address@hidden gnu/packages/bootstrap.scm:150 
>> 3d216c0>) ("gcc" #<package address@hidden gnu/packages/bootstrap.scm:150 
>> 3d21600>) ("binutils" #<package address@hidden 
>> gnu/packages/bootstrap.scm:150 3d21540>) ("coreutils&co" #<package 
>> address@hidden gnu/packages/bootstrap.scm:150 3d21480>) ("bash" #<package 
>> address@hidden gnu/packages/bootstrap.scm:150 3d21480>) ("guile" #<package 
>> address@hidden gnu/packages/bootstrap.scm:150 3d213c0>))
>>
>> This comes from the fact that the ‘inputs’ field is not overridden,
>> unlike in the case of ‘gnu-make-no-implicit-inputs’.
>>
>> To solve this, the solution is to add this one ‘inputs’ line:
>>
>> (define gnu-make-explicit-inputs
>>   (let ((p (package-with-explicit-inputs gnu-make
>>                                          (%bootstrap-inputs+toolchain)
>>                                          #:guile %bootstrap-guile)))
>>     (package-with-bootstrap-guile
>>      (package (inherit p)
>>               (name "make-explicit-inputs")
>>               (inputs '())                        ;<- HERE
>>               (arguments (package-arguments p))))))
>>
>> Perhaps you hit similar cases on ‘wip-bootstrap’?  It’s easy to leave
>> out too many inputs…
>
> I tried this!  The dependencies look OK, but the package won't build --
> there's no tar, make etc.

Ah, true!

> That can be fixed by repeating the explicit inputs, like this:
>
> (define gnu-make-explicit-inputs
>   (let ((p (package-with-explicit-inputs gnu-make
>                                          (%bootstrap-inputs+toolchain)
>                                          #:guile %bootstrap-guile)))
>     (package-with-bootstrap-guile
>      (package (inherit p)
>               (name "make-explicit-inputs")
>               (inputs (%bootstrap-inputs+toolchain))
>               (native-inputs '())
>               (arguments (package-arguments p))))))
>
> ...but that looks a bit strange: if we have to mention the inputs a
> second time the advantage over using the `gnu-make-no-implicit-inputs'
> package description becomes real small?

The key thing is that ‘package-with-explicit-inputs’ works recursively:
it adds (it does *not* replace) inputs to the whole package graph.

> I also tried
>
>      (inputs (package-inputs p))
>
> but that pulls in gcc-bootstrap-0 again; which lead me to believe
> `package-with-explicit-inputs' has no observable effect?

Consider this:

  (define x
    (let ((p (package-with-explicit-inputs gnu-make
                                           (%bootstrap-inputs+toolchain)
                                           …)))
      …))

Here ‘%bootstrap-inputs+toolchain’ is called from the top level, when
‘%current-system’ has its default value.  So if you’re on x86_64, you
get the x86_64 inputs.

So it’s not a bug per se, but it’s definitely an annoyance.

I just realized that there’s already a fix for this, which is to pass
‘package-with-explicit-inputs’ a procedure rather than the input list,
like this:

  (package-with-explicit-inputs gnu-make
                                %bootstrap-inputs+toolchain
                                …)

Does it work for you?

Thanks,
Ludo’.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]