[BusyBox] sh fails to compile under uClibc

[Mark Whitley]

On Thu, Mar 15, 2001 at 12:35:58PM -0700, Erik Andersen wrote:
> On Thu Mar 15, 2001 at 12:14:42PM -0700, Mark Whitley wrote:
> > On Thu, Mar 15, 2001 at 10:24:48AM -0800, Richard Doyle wrote:
> > > ../uClibc/extra/gcc-uClibc/gcc-uClibc-i386 -I- -Wall -Os -fomit-frame-po
> > > inter -D_GNU_SOURCE  -DBB_VER='"0.50pre"' -DBB_BT='"2001.03.15-18:29+000
> > > 0"' -DUSE_SYSTEM_PWD_GRP  -I.  -c sh.c -o sh.o
> > > sh.c: In function `expand_arguments':
> > > sh.c:1072: called object is not a function
> > > sh.c:1119: called object is not a function
> > > make: *** [sh.o] Error 1
> > > 
> > > Compile works fine with gcc. Using default Config.h, current snapshot
> > > from busybox.net, current uClibc, kernel 2.4.2-ac8.
> > 
> > Fix is in CVS. Please grab, please test.
> 
> Mark, I don't want this fix in there, please take it back out.  The problem is
> this is not a fix, and sh.c is not at fault.  The problem if that uClibc has a
> bug that needs fixing where it currently defines
> 
>     #define index strchr
>     #define rindex strrchr
> 
> This is wrong.  It needs to implement real func wrappers instead of #defines: 
> 
> char *index (const char *s, int c)
> {
>     return( strchr(s, c));
> }
> char *rindex (const char *s, int c)
> {
>     return( strrchr(s, c));
> }
> 
> We do not make uClibc better by adding workarounds to busybox,
> we make it better by fixing uClibc bugs in uClibc.  I was planning
> on doing this after the relase, but I'll fix uClibc now since folks
> are getting anxious.

Yoiks! Okay, I have reversed the change. Are you going to make the func
wrappers in uClibc or do you want me to do it?

Andreas Neuhaus (and doubtless others) are interested in this, so I'm cc'ing
the mailing list.


Mark Whitley
markw at lineo.com