Re: PING: [RFA/i386] 2 more patterns in i386_analyze_stack_align

[Joel Brobecker <brobecker@adacore.com>]

[-- Attachment #1: Type: text/plain, Size: 486 bytes --]

> Ah wait, that's true even for "absolute" code.  I had a quick look at
> the GCC code and it seems to always treat %ebx as callee-saved.  So
> I'm happy with leaving it out, but you could add a comment saying so
> to prevent us from having this same discussion in about two months
> ;-).

Sure! How does the attached patch look? Does it contain enough info?

2007-01-05  Joel Brobecker  <brobecker@adacore.com>

        * i386-tdep.c (i386_analyze_stack_align): Add comment.

-- 
Joel

[-- Attachment #2: comment.diff --]
[-- Type: text/plain, Size: 840 bytes --]

Index: i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.227
diff -u -p -r1.227 i386-tdep.c
--- i386-tdep.c	3 Jan 2007 19:01:25 -0000	1.227
+++ i386-tdep.c	5 Jan 2007 14:22:11 -0000
@@ -497,6 +497,10 @@ static CORE_ADDR
 i386_analyze_stack_align (CORE_ADDR pc, CORE_ADDR current_pc,
 			  struct i386_frame_cache *cache)
 {
+  /* The register used by the compiler to perform the stack re-alignment 
+     is, in order of preference, either %ecx, %edx, or %eax. GCC should
+     never use %ebx as it always treats it as callee-saved, whereas
+     the compiler can only use caller-saved registers.  */
   static const gdb_byte insns_ecx[10] = { 
     0x8d, 0x4c, 0x24, 0x04,	/* leal  4(%esp), %ecx */
     0x83, 0xe4, 0xf0,		/* andl  $-16, %esp */