Changes in the handling of comments caused some length fields to be used
in new ways.  The new way broke the existing adaptation for big endian
support; this patch repairs that adaptation.

Signed-off-by: Al Stone <ahs3@redhat.com>

Index: acpica-unix-20190405/source/compiler/aslcodegen.c
===================================================================
--- acpica-unix-20190405.orig/source/compiler/aslcodegen.c
+++ acpica-unix-20190405/source/compiler/aslcodegen.c
@@ -497,8 +497,7 @@ CgWriteTableHeader (
 
     /* Table length. Checksum zero for now, will rewrite later */
 
-    DWord = sizeof (ACPI_TABLE_HEADER) + Op->Asl.AmlSubtreeLength;
-    ACPI_MOVE_32_TO_32(&AslGbl_TableHeader.Length, &DWord);
+    AslGbl_TableHeader.Length = sizeof (ACPI_TABLE_HEADER) + Op->Asl.AmlSubtreeLength;
 
     /* Calculate the comment lengths for this definition block parseOp */
 
@@ -544,6 +543,8 @@ CgWriteTableHeader (
             CvDbgPrint ("    Length: %u\n", CommentLength);
         }
     }
+    DWord = AslGbl_TableHeader.Length;
+    ACPI_MOVE_32_TO_32(&AslGbl_TableHeader.Length, &DWord);
 
     AslGbl_TableHeader.Checksum = 0;
     Op->Asl.FinalAmlOffset = ftell (AslGbl_Files[ASL_FILE_AML_OUTPUT].Handle);