Date: Wed, 24 Sep 2008 22:20:03 GMT From: "Jeremy Messenger" <mezz7@cox.net> To: gnome@FreeBSD.org Subject: Re: ports/127297: Redefinition warnings in print/freetype2 Message-ID: <200809242220.m8OMK3G3061662@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR ports/127297; it has been noted by GNATS.
From: "Jeremy Messenger" <mezz7@cox.net>
To: "Jeremy Chadwick" <koitsu@freebsd.org>
Cc: bug-followup@freebsd.org
Subject: Re: ports/127297: Redefinition warnings in print/freetype2
Date: Wed, 24 Sep 2008 17:18:14 -0500
On Wed, 24 Sep 2008 17:07:20 -0500, Jeremy Chadwick <koitsu@freebsd.org>
wrote:
> On Wed, Sep 24, 2008 at 11:58:26AM -0500, Jeremy Messenger wrote:
>>> The pre-patch section comments out #define
>>> TT_CONFIG_OPTION_BYTECODE_INTERPRETER in
>>> include/freetype/config/ftoption.h (when WITHOUT_TTF_BYTECODE_ENABLED
>>> is
>>> not defined by the user), but then earlier in the Makefile, it
>>> re-enables it:
>>
>> I am not seeing any bug. When the WITHOUT_TTF_BYTECODE_ENABLED is not
>> defined then the TT_CONFIG_OPTION_BYTECODE_INTERPRETER will be enable,
>> which is a correct behavior. Unless I am blind?
>>
>> {snip}
>>
>> There is no problem. As for the redefine warning, I think it's harmless
>> since it's still enable with no change as it's merely enable again.
>
> The bug is in the port Makefile as I described. The CFLAGS addition of
> -DTT_CONFIG_OPTION_BYTECODE_INTERPRETER is 100% worthless, and induces
> the warnings.
>
> Let's step through each piece of the build process.
>
> Someone ultimately does "make" in ports/print/freetype2. They do not
> have WITHOUT_TTF_BYTECODE_ENABLED defined.
>
> In the port Makefile:
>
> .if !defined(WITHOUT_TTF_BYTECODE_ENABLED)
> CFLAGS+= -DTT_CONFIG_OPTION_BYTECODE_INTERPRETER
> .endif
>
> At this point TT_CONFIG_OPTION_BYTECODE_INTERPRETER will be defined
> during the build of the actual program. You can even see it being
> defined forcefully in "cc".
>
> Then later, in the port Makefile, we find this:
>
> .if !defined(WITHOUT_TTF_BYTECODE_ENABLED)
> @${REINPLACE_CMD} -e '/#define
> TT_CONFIG_OPTION_BYTECODE_INTERPRETER/s,/\* \(.*\) \*/,\1,; /#define
> TT_CONFIG_OPTION_UNPATENTED_HINTING/s,^\(.*\),/* \1 \*/,' \
> ${WRKSRC}/include/freetype/config/ftoption.h
> .endif
>
> The $REINPLACE_CMD code never gets executed, which makes sense because
> WITHOUT_TTF_BYTECODE_ENABLED is not defined.
Incorrect. See that '!', so the REINPLACE_CMD does get executed. It's why
I have shown you my example.
> So let's see what include/freetype/config/ftoption.h contains shall we?
>
> $ egrep '(#define|#undef) TT_CONFIG_OPTION_BYTECODE_INTERPRETER'
> work/freetype-2.3.7/include/freetype/config/ftoption.h
> #define TT_CONFIG_OPTION_BYTECODE_INTERPRETER
>
> Oh look, it's being #define'd in the .h file already!
>
> So what's the purpose of the CFLAGS part? To induce warnings? :-)
> Seeing this per *every C file* for no good reason is absurd:
>
> In file included from ./builds/unix/ftconfig.h:42, from
> /usr/ports/print/freetype2/work/freetype-2.3.7/include/freetype/freetype.h:41,
> from
> /usr/ports/print/freetype2/work/freetype-2.3.7/include/freetype/ftmodapi.h:24,
> from
> /usr/ports/print/freetype2/work/freetype-2.3.7/include/freetype/ftrender.h:24,
> from
> /usr/ports/print/freetype2/work/freetype-2.3.7/include/freetype/internal/ftobjs.h:30,
> from
> /usr/ports/print/freetype2/work/freetype-2.3.7/src/smooth/ftgrays.c:141,
> from
> /usr/ports/print/freetype2/work/freetype-2.3.7/src/smooth/smooth.c:22:
> /usr/ports/print/freetype2/work/freetype-2.3.7/include/freetype/config/ftoption.h:462:1:
> warning: "TT_CONFIG_OPTION_BYTECODE_INTERPRETER" redefined
> <command-line>: warning: this is the location of the previous definition
>
> I'm begging you to try it for yourself. Remove the CFLAGS modifier
> portion of the Makefile and then:
>
> # cd /usr/ports/print/freetype2
> # make clean
> # make
>
> You won't see any warnings.
>
> And try it with WITHOUT_TTF_BYTECODE_ENABLED=true too:
>
> # cd /usr/ports/print/freetype2
> # make clean
> # make WITHOUT_TTF_BYTECODE_ENABLED=true
>
> No warnings, and TT_CONFIG_OPTION_BYTECODE_INTERPRETER will be
> commented out in the .h file (as expected), because the $REINPLACE_CMD
> code will take care of it.
>
> Please reconsider your "there is no bug here" comment. There is most
> definitely a bug -- the FreeBSD port Makefile is causing tons of
> unnecessary warnings for *no good reason*.
The warnings don't mean it's bug.
Cheers,
Mezz
--
mezz7@cox.net - mezz@FreeBSD.org
FreeBSD GNOME Team
http://www.FreeBSD.org/gnome/ - gnome@FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200809242220.m8OMK3G3061662>