svn commit: r457245 - in head/net-p2p: bitcoin litecoin

[Jan Beich]

Alexey Dokuchaev <danfe at FreeBSD.org> writes:

> On Mon, Dec 25, 2017 at 06:07:48PM +0000, Jan Beich wrote:
>
>> New Revision: 457245
>> URL: https://svnweb.freebsd.org/changeset/ports/457245
>> 
>> Log:
>>   net-p2p/bitcoin: unbreak with boost 1.66
>>   
>>   PR:		224129 224132
>>   Obtained from:	upstream (Bitcoin 0.16.0)
>>   Approved by:	portmgr blanket
>> 
>> Modified: head/net-p2p/bitcoin/Makefile
>> ...
>> @@ -7,6 +7,9 @@ DISTVERSIONPREFIX=	v
>>  PORTREVISION?=	2
>>  CATEGORIES=	net-p2p finance
>>  
>> +PATCH_SITES=	https://github.com/${GH_ACCOUNT}/${GH_PROJECT}/commit/
>> +PATCHFILES=	1ec0c0a01c31.patch:-p1
> ...
>> 
>> Modified: head/net-p2p/bitcoin/distinfo
>> ...
>> +SIZE (1ec0c0a01c31.patch) = 3535
>
> I'd like to bring to your attention (again) that this approach of fetching
> external patches is fragile and already had broken things in the past.

Well, I did forget but
- PATCHFILES is less error prone, saving me accidental embarrasment
- PATCHFILES sends a strong message there were no modifications
  downstream which maybe important for fixes under blanket against
  security-sensitive software
- PATCHFILES is as mirror-(un)friendly as _GH0 suffix'd distfiles
- USE_GITHUB can be extended to to handle PATCHFILES to establish
  a better declarative style, decrease the number of clumsy commits

r454171 review overacted on the format change, as if GitHub distfiles
never change due to moved tags. The traditional way of backporting
upstream patches was never standardised.

> Putting 3535-byte patch file in ${FILESDIR} is perfectly fine for its size
> (and EXTRA_PATCHES also supports :-p1).

Unconditional EXTRA_PATCHES is a style bug. PHB defines EXTRA_PATCHES
only for conditional usage. Maybe portlint should warn about this.

--
Let's color the bikeshed as green.