Re: Update mafft to 7.245.

[Ricardo Wurmus]

Ben Woodcroft <address@hidden> writes:

> On 17/12/15 22:47, Ricardo Wurmus wrote:
>> Ben Woodcroft <address@hidden> writes:
>>
>>>> What do you think?
>>> I think you are quite valiant. As I say, I cannot be confident in my
>>> testing (even those in the readme). The diff looked mostly fine by eye,
>>> but there was some issues near the end which mangled things somewhat
>> What was mangled?
> Original:
>
> tmpawk=`which nawk 2>/dev/null | awk '{print $1}'`
> if [ -x "$tmpawk" ]; then
>          prog="$tmpawk"
> fi
>
> tmpawk=`which gawk 2>/dev/null | awk '{print $1}'`
> if [ -x "$tmpawk" ]; then
>          prog="$tmpawk"
> fi
>
>
> After your patch:
>
> tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk=`which 
> n/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk 
> 2>/dev/null | /gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/aw
> if [ -x 
> "$tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk" ]; then
> prog="$tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk"
> fi
>
> tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk=`which 
> g/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk 
> 2>/dev/null | /gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/aw
> if [ -x 
> "$tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk" ]; then
> prog="$tmp/gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk"
> fi

Oh, yes, that’s terrible.  Thanks for catching this!  (I don’t see why
my regex would do that, but I’m too tired to check this now.)

> The most recent patch:
>
> tmpawk=`which nawk 2>/dev/null | 
> /gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk '{print $1}'`
> if [ -x "$tmpawk" ]; then
>          prog="$tmpawk"
> fi
>
> tmpawk=`which gawk 2>/dev/null | 
> /gnu/store/k8qgvgwn5anbfy8r70h938kxgd46cyxx-gawk-4.1.3/bin/awk '{print $1}'`
> if [ -x "$tmpawk" ]; then
>          prog="$tmpawk"
> fi

Nice!

>> Other than that the patch does look fine.  If you confirm that this is
>> what you intended then I’ll push it as is.
> Thanks, if you are happy. This was just supposed to be a simple update..

Sorry for the delay!  It just never feels good to me to propagate
inputs.  If ever this can be avoided with a bit of patching I’d like to
try that first.

> I tried adding a check procedure but this didn't work: mafft refused to 
> run, when it runs just fine from the store. I was loath to debug that. 
> Instead, I was wondering if there was a way to test after installation? 
> If these tests could be run in a container that excluded native-inputs 
> (but perhaps some extra test-specific dependencies if required), then I 
> think such a procedure could be generally quite useful. It would catch 
> the errors I made in the original patch, for instance.

You could reorder the phases such that the check phase runs after
installation.  We do this for some Python packages as well.  It’s just a
matter of

   (delete 'check)
   (add-after 'install 'check
     (lambda ...))

Do you want to give this a try or shall I just apply the latest patch
you sent, leaving this for some time later?

Thank you for your patience!

~~ Ricardo