Gunther Nikl gni@gecko.de
Thu Sep 18 07:42:00 GMT 2003

On Wed, Sep 17, 2003 at 06:40:16PM +0200, Bernardo Innocenti wrote:
> >   "%{m68030:%{!ansi:%{!std=*:-Dmc68030}%{std=gnu*:-Dmc68030}}" \
> >   " -D__mc68030 -D__mc68030__} "
> This would defeat the whole point of the patch, which was getting rid
> of CPP_SPEC as requested by Richard. CPP_SPEC has been obsoleted and
> replaced by the TARGET_XXX_CPP_BUILTINS macros.

  Are you sure that CPP_SPEC is obsolete? Maybe you are thinking of
  CPP_PREDEFINES? Wasting a target bit only to define a symbol is a
  bad idea, IMHO.

> >>And no one complained about that so far, uh?
> >
> >  Maybe nobody uses my port ;-(
> Isn't your GCC port in Geek Gadgets? Oh, wait... I'm quite uninformed
> about the latest Amiga news. Does Geek Gadgets still exist?

  You can still download sources, diffs and binary packages.

> By the way, don't you want any of your patches in mainline? Time is running
> out... According to Mark Mitchell, GCC will switch to stage 3 very soon.

  Without copyright assignment? Of course I would like to see the port in
  mainline but I fear there are to many issues how the port operates and
  I probably can't address them. Currently I am stuck to find the cause
  for code generation bug which is triggered by -fstrength-reduce for
  -fbaserel (a pic mode). I tried to find the cause for the bug but I don't
  know if thats a real GCC bug or a bug in the port.

> it should never happen that an FP instruction gets emitted when TARGET_68881
> is off.

  I agree, but TARGET_68881 is true for -m680[46]0.

> I think the NetBSD/m68k kernel relies on it to avoid the burden of saving
> and restoring FP registers in syscalls and interrupts.

  NetBSD uses -msoft-float which turns FPU instructions off.

> If it's really needed for compatibility, I could add yet another target
> flag such as TARGET_SOFTFP and use that.

  Please don't waste a target bit for this. If the old behaviour was a bug,
  then fix it.


More information about the Gcc-patches mailing list