Re: [PATCH] unifdef: teach it about defined(FOO) syntax

[Sam Ravnborg <sam@ravnborg.org>]

On Fri, Jun 26, 2009 at 03:04:53PM -0700, Andrew Morton wrote:
> On Sun, 21 Jun 2009 20:21:32 +0200
> Denys Vlasenko <vda.linux@googlemail.com> wrote:
> 
> > Hi,
> > 
> > uclibc project patched unifdef.c (which we stole from kernel)
> > so that it understands defined(FOO) in addition to defined FOO,
> > and also taught it to understand short-circuited evaluation
> > of && and ||
> > 
> > The patch is attached. (Sorry, not inline, I fear Google
> > web mail interface may mangle it).
> > 
> > I ran these commands in unpatched and patched tree:
> > 
> > make ARCH=i386 CROSS_COMPILE=i486-linux-uclibc- defconfig
> > ln -s asm-x86 include/asm
> > make ARCH=i386 CROSS_COMPILE=i486-linux-uclibc- headers_install
> > 
> > and then diffed usr/*. The difference clearly shows that
> > new unifdef works better than old one:
> > 
> > 
> > linux-2.6.30.test/usr/include/linux/acct.h:
> > @@ -59,9 +59,7 @@ struct acct
> >  	comp_t		ac_majflt;		/* Major Pagefaults */
> >  	comp_t		ac_swaps;		/* Number of Swaps */
> >  /* m68k had no padding here. */
> > -#if !defined(CONFIG_M68K) || !defined(__KERNEL__)
> >  	__u16		ac_ahz;			/* AHZ */
> > -#endif
> >  	__u32		ac_exitcode;		/* Exitcode */
> >  	char		ac_comm[ACCT_COMM + 1];	/* Command Name */
> >  	__u8		ac_etime_hi;		/* Elapsed Time MSB */
> > 
> > 
> > linux-2.6.30.test/usr/include/linux/soundcard.h:
> > @@ -1033,7 +1033,6 @@ typedef struct mixer_vol_table {
> >   */
> >  #define LOCL_STARTAUDIO		1
> > 
> > -#if !defined(__KERNEL__) || defined(USE_SEQ_MACROS)
> >  /*
> >   *	Some convenience macros to simplify programming of the
> >   *	/dev/sequencer interface
> > @@ -1275,4 +1274,3 @@ void seqbuf_dump(void);	/* This function
> >  		(SEQ_DUMPBUF(), write(seqfd, (char*)(patchx), len))
> > 
> >  #endif
> > -#endif
> > 
> > ...
> >
> 
> eek.  Please avoid quoting anything which looks like a patch in the
> changelog.  Because when some poor schmuck comes along and tries to
> apply it, patch(1) goes sniffing around in the changelog, thinks it
> sees a patch and makes a huge mess of everything.
> 
> 
> > Please apply.
> 
> Unfortunately unifdef.c got changed after 2.6.30 and this patch throws
> more rejects than I am comfortable about fixing.

Russell King actually implemented the same functionality (more or less).
Denys - could you try if the one in the kernel is better/worse
then the one you have here.

The implementation Russell did looks much simpler than this
patch and so far it has only been a win.

	Sam