[BusyBox] [BB-1.0pre5][Diet-CVS] Busybox vs. diet: multiple build failures

[Manuel Novoa III]

Hello,

On Thu, Jan 15, 2004 at 05:23:10PM +0300, Samium Gromoff wrote:
> diet /opt/elcore/bin/mipsel-linux-gcc -I../include -Wall -Wstrict-prototypes -Wshadow -Os -fomit-frame-pointer -D_GNU_SOURCE -DNDEBUG     -DL_bb_vfprintf -c ../libbb/printf.c -o ../libbb/bb_vfprintf.o
> In file included from ../libbb/printf.c:47:
> ../include/libbb.h:299: warning: `struct sockaddr_in' declared inside parameter list
> ../include/libbb.h:299: warning: its scope is only this definition or declaration, which is probably not what you want
> ../include/libbb.h:301: warning: `struct sockaddr_in' declared inside parameter list
> ../libbb/printf.c:119:2: #error dietlibc is currently not supported. Please see the commented source.
> ../libbb/printf.c: In function `bb_vfprintf':
> ../libbb/printf.c:136: warning: implicit declaration of function `SET_FERROR_UNLOCKED'
> make: *** [../libbb/bb_vfprintf.o] Error 1
> 
> Nice. Killing the #error...
> 
....
> 
> _and_ i was unable to find a useful __stdio_file definition to form a
> meaningful SET_FERROR_UNLOCKED() macro for busybox.

>From a message to Felix on Sept 14, 2003:

  The entire purpose of libbb/printf.c is to make sure that output
  errors are checked.  Most people fail to check for error returns
  on printf, as they are so rare.  But providing some wrappers for
  the various *printf functions which set the stream error indicator
  when a printf error does occur, a number of busybox applets were
  made more reliable without undo increase in code size.  They simply
  needed to check ferror() on the streams they used prior to exit.

>From a message to Felix on Nov 13, 2003:

  Oh... I'm still waiting for a patch from you to implement the needed
  functionality so that the #error in the busybox code can be removed.
  It would be good to get that in before the 1.0 release.


At some point in between, I had explained what needed to be done and
suggested that he define a macro to insulate busybox from changes and
prevent problems if he upgraded his stdio implementation.  Can't find
that email at the moment.

In any case, no one connected with dietlibc has ever taken the time to
do it, in spite of the 'annoyance' recorded in the dietlibc 0.23 release
changelog.

Manuel