svn commit: r442588 - in head/www: nginx nginx-full

[Gerald Pfeifer]

On Thu, 8 Jun 2017, Bryan Drewery wrote:
> On 6/6/17 5:39 AM, Alexey Dokuchaev wrote:
>> It seems that everyone bumps port revisions whenever they please 
>> these days;

I can confirm I am neither doing nor omitting PORTREVISION bumps 
lightheartedly, Alexey. :-}

>> Just an exampler: r442562, where it was bumped for pkg-descr change 
>> (sic!) in a port that takes considerable time to build. :-(
> pkg-descr is part of the generated package. Bumping PORTREVISION in that
> commit had a real impact and change.  Plus the actual content change is
> not a small spelling change.

Thanks, Bryan!

> It's explaining something to users about the proper use of the port and
> how to obtain the right package for the right environment.
> 
> Yes it hurts to rebuild a port for something like this, but it is
> absolutely correct.

I find bumping (or not) PORTREVISION actually a bit tricky.

While I believe the case above is one where this is the right thing,
I have been skipping it for other cases where technically a bump would 
be required.  

Consider revision 443791:

Index: gcc5/pkg-descr
===================================================================
-WWW: http://gcc.gnu.org/
+WWW: https://gcc.gnu.org

Since both the old and new URLs work just perfectly fine, and the
former only redirects to the latter, and given that lang/gcc* are
somewhat heavier ports, in this case I opted not to do the bump.

Strictly speaking, however, it still would be applicable and when
I don't do it always kind of wait for someone to complain.  Of course,
when I then do bump in a stronger case, someone complains, too.  I 
guess you cannot win (until we have stronger tooling as Bryan hinted
at). ;-)

Gerald