Two reviews in record time, nice.

On 10/11/15 23:12, Efraim Flashner wrote:
> On Tue, 10 Nov 2015 22:18:10 +1000
> Ben Woodcroft <b.woodcroft@uq.edu.au> wrote:
>
>> Also had to fix the inputs. Hard not to notice these things in the
>> environment container - worked well thanks.
> Do they need to be propagated inputs? Did you try them as native-inputs?
> Often that's enough to take care of it.
The main program 'mafft' is actually a reasonably long shell script, 
which itself calls awk, grep, perl, etc. collectively many times (>100 I 
would guess). I think it is intended to be run where these programs are 
available. I could do as Ricardo suggests and run substitute* but this 
seems a bit error-prone and not very future-proof to me, especially when 
the shell script is difficult to exhaustively test. WDYT?

Note that mafft is probably not going to remain as a leaf because a 
number of bioinformatic tools not yet packaged rely on it.
> Also, instead of:
> Fix inputs.
> the commit message is more along the lines of:
>    [inputs]: Move inputs to propagated-inputs and add missing dependencies.
> but I normally have to check against older commit messages to see what others
> are doing and I try to copy that.
Hopefully the attached is a better attempt? I also added another small 
improvement, not distributing the manpage of a program that is excluded 
from distribution.

Thanks.