[980122]: Patch for new bug in libgcc2 for m68k-motorola-sysv

manfred@s-direktnet.de manfred@s-direktnet.de
Sun Jan 25 23:23:00 GMT 1998


On Sun, 25 January 1998, 20:52:04, law@cygnus.com wrote:

 > 
 >   In message < 199801251808.TAA27564@saturn.s-direktnet.de >you write:
 >   > The introduction of the __clear_insn_cache function for this particular
 >   > system inherited with gcc-2.8.0's merge causes bootstrap failures for me.
 >   > The new code calls the function `memctl' which isn't defined in any of
 >   > the libc's on this system!
 >   > 
 >   > As my system is pretty much original, I actually don't know, how Philippe
 >   > <phdm@macqel.be> got this change to work on his machine - Philippe, comment  > s?
 >   > 
 >   > To fix it, I added a check for memctl to configure.in and enabled the new
 >   > code only if HAVE_MEMCTL is defined.
 > Actually, this is a target issue, and thus is not suitable for autoconf
 > (it happens to work in your case because host == target when doing a native
 > build).
 > 
 > So, the right thing to do is determine what systems need it and put that
 > info in the target .h files statically.
 > 
 > Though as you mention, I'd like to know how Pilippe got this change to work
 > on his machine.  Maybe it's due to a different version of the OS.

This is what  Philippe sent to my business  e-mail address (I left too
early last Friday :-):

On Fri, 23 January 1998, 16:38:02, phdm@macqel.be wrote:

 > > 
 > > On my system _NO_ include file from /usr/include does define MCT_TEXT, and
 > > _NO_ libc.a (libc881.a, ...) does define the memctl function, either :-(
 > 
 > Sorry about that.  I had noticed that you do not have a 68040, as your check
 > results never fail on nested functions testcases, but I did not suspect
 > that you could'nt compile my patch.
 > 
 > > 
 > > I fixed this problem for me by adding a check for memctl to configure(.in),
 > > and using an empty __clear_insn_cache body for #ifndef HAVE_MEMCTL.
 > 
 > You should rather test #ifdef MCT_TEXT in libgcc2.c, and not touch
 > configure(.in).

I'll send a different patch checking only the MCT_TEXT as Philippe suggested.

 > 
 > > 
 > > BUT, what particular OS version are you using? Where do you get the define
 > > and the function from?
 > > 
 > sudm/mac_tst cat /usr/options/BOS
 > BOS
 > SYSTEM V/68 Base Operating System R3V7
 > FE03.71 RM04
 > Fri Apr  2 11:02:59 1993

My one is:

$ cat /usr/options/BOS 
BOS
SYSTEM V/68 Base Operating System R3V7
FE03.70 RM03
Mon Nov 11 16:31:33 1991

What a difference a cent can make...

Manfred



More information about the Gcc-bugs mailing list