Bootstrap failure on sparc-linux-gnu
Laurent GUERBY
laurent@guerby.net
Thu Apr 22 16:35:00 GMT 2004
On Thu, 2004-04-22 at 17:48, Eric Botcazou wrote:
> A starting point would be to look at what dwarf2out.c is passed in both
> cases.
A wild guess in the same vein as my 2001-10-27 change:
I looked at trans.c 1.50 vs 1.51 and this is the only place
I could spot where more than one function is called
is an argument list (a potential source of non-determinism).
Could someone try this out on a platform where the problem occurs?
Laurent
2004-04-22 Laurent Guerby <laurent@guerby.net>
* trans.c (gigi): Fix non determinism leading to bootstrap
comparison failures.
Index: trans.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/ada/trans.c,v
retrieving revision 1.52
diff -w -u -r1.52 trans.c
--- trans.c 19 Apr 2004 15:20:14 -0000 1.52
+++ trans.c 22 Apr 2004 16:08:55 -0000
@@ -2112,8 +2112,11 @@
for (gnat_temp = First (Elsif_Parts (gnat_node));
Present (gnat_temp); gnat_temp = Next (gnat_temp))
{
- tree gnu_elseif
- = build_nt (IF_STMT, gnat_to_gnu (Condition (gnat_temp)),
+ tree gnu_cond, gnu_elseif;
+
+ gnu_cond = gnat_to_gnu (Condition (gnat_temp))
+ gnu_elseif
+ = build_nt (IF_STMT, gnu_cond,
build_block_stmt (Then_Statements (gnat_temp)),
NULL_TREE, NULL_TREE);
@@ -2123,10 +2126,18 @@
gnu_result = gnu_elseif;
}
- gnu_result = build_nt (IF_STMT, gnat_to_gnu (Condition (gnat_node)),
- build_block_stmt (Then_Statements (gnat_node)),
+ {
+ tree gnu_cond, then_block, else_block;
+
+ gnu_cond = gnat_to_gnu (Condition (gnat_node));
+ then_block = build_block_stmt (Then_Statements (gnat_node));
+ else_block = build_block_stmt (Else_Statements (gnat_node));
+
+ gnu_result = build_nt (IF_STMT, gnu_cond,
+ then_block
nreverse (gnu_result),
- build_block_stmt (Else_Statements (gnat_node)));
+ else_block);
+ }
break;
case N_Case_Statement:
More information about the Gcc
mailing list