Re: grx23 under as-2.95 - Grx

9 Sep 1999


      Sorry, pressed send button too early in previous mail)
Now about patch. It seems to be Ok. I had to do the same when tried 
to build Allegro using recent binutils snapshot.
Andris
On 8 Sep 99, at 22:29, Ian Miller wrote:
...
Ahem.
I'm using pretty much up to date binutils 2.95
right now. I wouldn't recommend it unless it
doesn't matter to you much if your software
doesn't work. That said, they do seem to
work fine. However,...
as 2.95 interprets a couple of GRX inline
assembler call instructions as
PC-relative and chokes on them.
as 2.81 interpreted them as absolute
calls, and this is evidently what the code
intends.
The following simple patch, duplicated at
http://www.shelob.force9.co.uk/djgpp/grxasm.dif,
applies the * prefix to the call absolute
operands as the as 2.95 manual requires.
It works for me.
IF IN DOUBT, DO NOT APPLY THIS PATCH
Happy hacking,
Ian
--
Ian Miller, Dorset, UK
*** src/vdrivers/vesa_pm.orig.c Mon May 11 20:20:28 1998
--- src/vdrivers/vesa_pm.c Wed Sep  8 21:41:52 1999

*** 349,355 ****
  static INLINE void PM_banking(short BX, short DX) {
    __asm__ volatile (
      " pushal   \n"
!     " call %3  \n"
      " popal      "
      : /* no output */
      : "a" (0x4F05), "b" (BX), "d" (DX),
--- 349,355 ----
  static INLINE void PM_banking(short BX, short DX) {
    __asm__ volatile (
      " pushal   \n"
!     " call *%3  \n"
      " popal      "
      : /* no output */
      : "a" (0x4F05), "b" (BX), "d" (DX),

*** 362,368 ****
      " pushal             \n"
      " movw %%ax, %%es    \n"
      " movw $0x4f05, %%ax \n"
!     " call %3            \n"
      " popal                "
      : /* no output */
      : "a" (es), "b" (BX), "d" (DX),
--- 362,368 ----
      " pushal             \n"
      " movw %%ax, %%es    \n"
      " movw $0x4f05, %%ax \n"
!     " call *%3            \n"
      " popal                "
      : /* no output */
      : "a" (es), "b" (BX), "d" (DX),