[uClibc]Re: latest diff vs. my uClibc tree

[Erik Andersen]

On Sat Jan 06, 2001 at 09:58:45PM -0600, Manuel Novoa III wrote:
> 
> Erik,
> 
> Below is the diff from my current tree.  Some of the changes require
> changes in the arm and m68k start files.  See the _start.c file for i386
> w.r.t. libc_init.
> 
> Summary of changes:
> 
> Fixed general Makefile problem of dependencies with COBJS.
> Added unified syscalls for i386 (like dietlibc).  Saves about 2.8k.
> Added brk and sbrk for i386
> Added the option of using malloc-930716 from libc-5.3.12. (requires sbrk)
> Added a wrapper for gcc to ease use of uClibc.
>     "make CC=gcc-uClibc-i386"  works to build busybox with uClibc.

cool

> Defined NDEBUG in CFLAGS if not debugging.
> Added gnu getopt from libc-5.3.12.  Using weak symbols, this is only
>     linked what long options are needed.  The smaller original getopt
>     from uClibc is used otherwise.

While you are there -- there is a long standing bug in libc5 getopt...
It does not set the global var optind=1 until afer getopt has been called.

> Using weak symbols, atexit is only linked if needed.
> Removed onexit as it varied in behavior from gnu libc.
> Replaced broken (and large) qsort.c with a small shell sort.  Saves over 1k.

cool

> Disable fprintf in system.c.  Was it there for debugging?

must have been -- I don't recall putting it there though.

> Fixed checking for strerror.c and strsignal.c.  Added testing in test/string.
> Fixed bug in net/resolve.c (fd possibly not initialized).
> Changes to stdio.c and stdio.h.  (initialization, buffers, FILEs)
> All *s*printf functions now share the fake FILE.

k

> 
> Requires changes for arm and m68k _start files.
> Stdio now uses weak symbols to automate initialization. 

Cool.  I'll take a look and baring any wierdness it should
show up in CVS presently,

 -Erik

--
Erik B. Andersen   email:  andersen at lineo.com
--This message was written using 73% post-consumer electrons--