svn commit: r542249 - in head: devel/msgpack math/calc security/gnupg security/gnupg1 sysutils/z

[Adam Weinberger]

> On Jul 15, 2020, at 01:32, Alexey Dokuchaev <danfe at freebsd.org> wrote:
> 
> On Wed, Jul 15, 2020 at 07:28:28AM +0000, Adam Weinberger wrote:
>> New Revision: 542249
>> URL: https://svnweb.freebsd.org/changeset/ports/542249
>> 
>> Log:
>>  Unmute some build commands
>> 
>> @@ -30,7 +30,7 @@ USE_GCC=    yes
>> 
>> post-patch:
>> # GoogleTest doesn't support C++98 consumers since 1.10.0
>> -    @${REINPLACE_CMD} -e 's/c++98/c++11/' ${WRKSRC}/CMakeLists.txt
>> +    ${REINPLACE_CMD} -e 's/c++98/c++11/' ${WRKSRC}/CMakeLists.txt
> 
> This is not the build, this is patch command.  In the ports, patching
> traditionally happens silently.
> 
>> post-install:
>> -    @${STRIP_CMD} ${STAGEDIR}${PREFIX}/bin/calc \
>> +    ${STRIP_CMD} ${STAGEDIR}${PREFIX}/bin/calc \
>>        ${STAGEDIR}${PREFIX}/lib/lib*.so
> 
> Stripping also typically happens silenty.
> 
>> -    @${TOUCH} ${WRKSRC}/doc/*.texi
>> +    ${TOUCH} ${WRKSRC}/doc/*.texi
> 
> Touching dummy files (cookies) definitely should happen silently.
> 
>> post-install:
>> -    @${MV} ${STAGEDIR}${DATADIR}/help*.txt ${STAGEDIR}${DOCSDIR}
>> +    ${MV} ${STAGEDIR}${DATADIR}/help*.txt ${STAGEDIR}${DOCSDIR}
> 
> This one is correct.
> 
>> post-install-DOCS-on:
>> -    @${MKDIR} ${STAGEDIR}${DOCSDIR}
>> +    ${MKDIR} ${STAGEDIR}${DOCSDIR}
>> 
>> do-install:
>> -    @${MKDIR} ${STAGEDIR}${DATADIR}
>> +    ${MKDIR} ${STAGEDIR}${DATADIR}
> 
> We usually mute mkdir's.  Please do not "fix" what ain't broken, this
> just creates needless repochurn, pessimizes "svn blame" and solves
> no real problem, but just makes things less consistent throughout the
> ports tree.

No. Just because we all drew an arbitrary line in the sand at one point does not make it viable today. Those distinctions were made a long time ago when (a) everybody who wanted third-party software had to experience it; (b) end-users had to trouble-shoot build failures (in which case minimizing output was critical); and (c) we wanted the building/installing experience to be clean and quiet and simple.

Now it is far more useful for ports to show exactly what it’s doing. As an example, your “definitely should happen silently” example is a case where we do something that explicitly causes the build to regenerate something that it wouldn’t normally.

“Consistency” when each build output is completely different is not a particularly compelling argument for me. It used to be, but now I’m much more interested in preserving build output whenever possible.

# Adam


—
Adam Weinberger
adamw at adamw.org
https://www.adamw.org