TRTE / TRTRE fix - endian issue

[JamesWekel]

Fish,

Here is my proposed fix for the big-endian issue. I've also made one other small fix for a potential page cross on the initial MADDRL address lookup for the operand argument.

Jim

[Fish-Git]

Please see my comment in the GitHub Issues thread:

• Tests "TRTE/TRTRE-01-basic" fail on s390x CentOS 7 #528 (comment)

You need to fix your other CSWAP16 bugs that are causing Sparc to crash.

[Fish-Git]

FYI:  Here's (hopefully!) the Sparc fix:

--- hyperion-TRTE-Fix-1/general2.c	2022-11-30 16:03:32.508006500 -0800
+++ hyperion-TRTE-Fix-0/general2.c	2022-11-30 15:59:26.721574900 -0800
@@ -3258,9 +3258,7 @@
                 arg_ch  =  (temp_h << 8) | temp_l;
             }
             else
-            {
-                arg_ch = CSWAP16( *(U16*) buf_main_addr );
-            }
+                arg_ch = fetch_hw( buf_main_addr );
         }
         else
         {
@@ -3290,7 +3288,10 @@
                     fc      =  (temp_h << 8) | temp_l;
                 }
                 else
-                    fc =  CSWAP16( *(U16*) (fct_main_page_addr[ fc_page_no - fct_page_no ] + (fc_addr & PAGEFRAME_BYTEMASK)));
+                {
+                    BYTE* fct_main = fct_main_page_addr[ fc_page_no - fct_page_no ] + (fc_addr & PAGEFRAME_BYTEMASK);
+                    fc = fetch_hw( fct_main );
+                }
             }
             else
             {

[JamesWekel]

Fish,

As I was already building a big-endian U16 across a page boundary, I thought the code read better just continuing building the U16 value on the 'else' block.

As for copying the FC Table, I was trying to avoid that because the table could be 128 KB.

Hope this works on the Sparc 64. I've never worked on this hardware platform so understanding that there was an alignment issue with CSWAP16 was way beyond my pay grade. Still learning.  :-)

Jim

[Fish-Git]

Looks good to me! Pull Request accepted!