Ricardo Wurmus <rek...@elephly.net> skribis:
>>>>> +(define-public libiberty
>>>>> + (make-libiberty gcc))
>>>>
>>>> s/gcc/gcc-4.8/ if 4.8 is the last package installing it.
>>>
>>> I wanted to provide a default libiberty package for the default “gcc”.
>>> Otherwise any change to the default GCC version would require packages
>>> to update their libiberty input.
>>>
>>> GCC 4.8 actually installs libiberty somewhere in the “lib” output of
>>> gcc-4.8, but GCC 4.9 does not, nor does GCC 5.1. Is there a problem
>>> with “libiberty” as defined above, following whatever GCC version is set
>>> as the default?
>>
>> Oh you mean that 4.9 and 5.1 install it as well, just not in the “lib”
>> output, right?
>
> No, I mean that although 4.9 and 5.1 include the sources of libiberty
> neither of them install it as a library in *any* of their outputs.
>
> With 4.8 a package depending on libiberty could add
>
> ("gcc" ,gcc-4.8 "lib")
>
> as an input, but with 4.9 and 5.1 there does not seem to be any way to
> get libiberty but to explicitly install it. This is what the new
> “libiberty” package is supposed to do.
OK.
> Following the GCC version automatically dependent on whatever value the
> variable "gcc" has is just a bonus to simplify upgrades to the default
> version of GCC (as the recent move from 4.8 to 4.9).
I don’t get it. If we use ‘gcc’ instead of ‘gcc-4.8’, it doesn’t work,
because ‘gcc’ is currently an alias for ‘gcc-4.9’.
This is why I suggested using ‘gcc-4.8’ explicitly: We know it’s the
last version that installs libiberty.
Thanks,
Ludo’.
