This is the mail archive of the libc-hacker@cygnus.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [kwzh@gnu.org (Karl Heuer)] glibc-2.1


   Date: Thu, 11 Feb 1999 16:35:31 -0500
   From: Zack Weinberg <zack@rabi.columbia.edu>

   On 11 Feb 1999 12:32:23 -0800, Ulrich Drepper wrote:

   >From: kwzh@gnu.org (Karl Heuer)
   >To: drepper@gnu.org
   >Subject: glibc-2.1
   >Content-Type: text
   >
   >>The glibc-2.1 distribution recommends compiling with egcs
   >>rather than gcc; rms says that this violates our policy,
   >>and hence I've removed it from ftp.gnu.org for now.
   >
   >Isn't that nice?

   - Has anyone tried to explain the compatibility issues to rms?

   - Mark had an analysis of the problem that sounded like we could
   arrange for proper compatibility while allowing for use of gcc 2.8,
   with some stub functions.

AFAIK, the only problem is that the backwards compatibility
functions `__deregister_frame', `__register_frame' and
`__register_frame_table' are in seperate module in gcc 2.8.1's
libgcc.a wheras with egcs they are in the same module as ones with a
`_info' suffix.  So referencing `__register_frame' somewhere in
`soinit' should do the trick.

However, I seriously doubt whether these backwards compatibility
functions are really needed.  AFAIK no officially released version of
gcc or egcs ever used `__register_frame' and friends (please correct
me if I'm wrong).

As long as people do not use excessive optimizations, or try to
compile ``I'm Linus Torvalds, so I can safely use
undocumented/experimental gcc features and expect that this code still
compiles with a future compiler''-code, gcc 2.8.1 is quite a stable C
compiler.

Mark



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]