Danny Milosavljevic writes:

> Hi,
>
> thanks for the patch!
>
> On Sat, 14 Jan 2017 20:56:52 +0200
> Theodoros Foradis <theodoros.for@openmailbox.org> wrote:
>
>> * gnu/packages/fpga.scm (icestorm)[arguments]: Modify phases
>> "fix-usr-local" to replace reference to /usr/local/share.
>> ---
>>  gnu/packages/fpga.scm | 9 +++++----
>>  1 file changed, 5 insertions(+), 4 deletions(-)
>> 
>> diff --git a/gnu/packages/fpga.scm b/gnu/packages/fpga.scm
>> index f65eae8..eb05a39 100644
>> --- a/gnu/packages/fpga.scm
>> +++ b/gnu/packages/fpga.scm
>> @@ -225,11 +225,12 @@ For synthesis, the compiler generates netlists in the desired format.")
>>                            (string-append "PREFIX=" (assoc-ref %outputs "out")))
>>         #:phases
>>          (modify-phases %standard-phases
>> -          (add-after 'unpack 'remove-usr-local
>> -            (lambda _
>> -              (substitute* "iceprog/Makefile"
>> +          (add-after 'unpack 'fix-usr-local
>> +            (lambda* (#:key outputs #:allow-other-keys)
>
>> +              (substitute* '("iceprog/Makefile"
>
> Patching this files changes nothing in the final result.
>
> The package is built in an isolated container anyway - it's not like it can access /usr/local while it's building.
>
> And if we did that everywhere we would have lots of substitution commands all over the place as noise - doing nothing useful, drowning out the important signal of actually useful substitutions. Therefore, I don't think we should do this.
>
> What do the others think?
>
>> "icebox/icebox_vlog.py")
>
> ... oops. This one I definitely agree with (the /usr/local/share one).

I should have probably added a different phase for the
"icebox/icebox_vlog.py" substitution, to make clear it's the only path
which is a run-time reference. The installed icebox_vlog, can't find the
chipdb.txt file, if we don't make the substitution.

I attach the modified patch, adding a new phase instead.

Regards,
-- 
Theodoros Foradis