Re: HEADS-UP: Upcoming move of WWW definitions in all ports

On 9/6/22 9:04 PM, Stefan Esser wrote:
> Am 06.09.22 um 20:31 schrieb Steve Kargl:
>> On Tue, Sep 06, 2022 at 08:09:24PM +0200, Stefan Esser wrote:
>>> PortMgr has just accepted a change that has been in the making for 
>>> some time:
>>>
>>> The move of the "WWW:" entries at the end of the ports' pkg-descr 
>>> files into
>>> the ports' Makefiles in the form of "WWW=" definitions.
>>>
>>> The WWW lines will be put into the maintainer block, immediately 
>>> following
>>> the MAINTAINER and COMMENT lines.
>>>
>>> If there is more than one WWW: line a pkg-descr file, one URL will be
>>> selected to be moved into the Makefile, the other URLs will get a label
>>> "See also:" instead of "WWW:".
>>>
>>
>> You've told us the "What" and suggested a "When".  So, what is
>> the "Why?"
>>
>> % find . -name pkg-descr -maxdepth 3 | wc -l
>>     29110
>> % find . -name Makefile -maxdepth 3 | wc -l
>>     30343
>>
>> Assuming 90% of the files are touched, 54000 files suffering cosmetic 
>> churn?
> 
> The idea to move the WWW entry from the pkg-descr to the Makefile has 
> come up
> multiple times in the past, back in 2015 for the first time that I know of,
> see https://reviews.freebsd.org/D2949.
It's really interesting to see my work being resurrected after nearly 7 
years. But yeah thanks for the hard work and I believe my task was 
easier than what you are doing or did. I believe you or portmgr@ has a 
better script to complete the task. In case you are planning to use my 
script I would advise doing in batches per category so far from my 
experience. But then again I believe there are more competent ones than 
me to take care of this. :D

> The question whether the churn caused by commits to 2 files of nearly 
> all ports
> has been discussed within the PortMgr team for some time. The decision 
> to go
> ahead has now been made, after all prerequisites have been met.
> 
> 
> The WWW field is the only parameter in the INDEX file or in package 
> manifests
> that is not derived from some variable in the ports' Makefiles.
> 
> It has become stale in quite a number of ports over time, one reason 
> might be
> that the pkg-descr files are seldom touched after the initial commit. 
> And many
> of the URLs did not follow the rules laid out in the porters' handbook.
> 
> The most relevant WWW: entry in each pkg-descr file should be in the 
> last line
> (if there are multiple WWW: lines) and this line is currently extracted 
> using
> awk when the INDEX is built. But in fact, many ports have the relevant 
> URL in
> some other line and thus do not get any useful URL in the INDEX.
> 
> Moving the WWW value into the Makefile will make it easier to maintain 
> and test
> whether it is still valid. It removes processing overhead from "make 
> index" and
> it is a chance to select the most relevant entry from multiple URLs in 
> many of
> the pkg-descr files.
> 
> Best regards, STefan
>