Re: Patches for FreeBSD 4.2

[Mark Kettenis <kettenis at science dot uva dot nl>]

Stu Grossman <grossman@juniper.net> writes:

> Here is a set of patches that fixes a bunch of problems (and some testsuite
> failures) for FreeBSD 4.2.

Thanks.  A few comments about the things that fall on my turf:

> + 	* i386-tdep.c (i386_store_return_value):  Make sure that floats get
> + 	stored at the top of the FP stack.

Already fixed :-).

> + 	* i386-freebsd-nat.c:  New file to support FreeBSD specific stuff.

* I'd prefer the name i386fbsd-nat.c, since we already have
  alphafbsd-tdep.c.

* This file won't compile on FreeBSD 3.x, since <machine/sigframe.h>
  doesn't exist.  Besides, the implementation of a *_frame_saved_pc
  function belongs in a target-dependent file, and shouldn't depend on
  the target headers.  What's wrong with the old implementation?

  Hmm, `struct sigcontext' has been changed.  The CVS logs in the
  FreeBSD tree seems to imply that it is possible to distinguish
  between a new and an old signal handler frame.  So perhaps we can
  distinguish between the two.  Otherwise we have to provide different
  implementations for old and new FreeBSD versions.

* I think the SOFTWARE_SINGLE_STEP_P stuff can be removed from your
  version of child_resume.

> + 	* config/i386/tm-fbsd.h:  Define FRAME_SAVED_PC to use FreeBSD specific
> + 	code to extract return PC from signal frames.
> + 	* Define FRAME_CHAIN_VALID to use func_frame_chain_valid, since ELF
> + 	doesn't generate N_TEXT object file markers.
> + 	* Remove SIGTRAMP_START and SIGTRAMP_END.  Move all of that logic into
> + 	a FreeBSD specific function invoke via IN_SIGTRAMP.

Why did you comment out USE_STRUCT_CONVENTION here?  AFAIK FreeBSD
uses the old GCC 1 conventions for passing short structs.

Mark