This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Some VRP improvements



On Jun 2, 2005, at 5:43 PM, Andrew Pinski wrote:



On Jun 2, 2005, at 4:56 PM, Diego Novillo wrote:
Yeah, exactly the same stunt that the Fortran FE was pulling on
me.

For some reason memcmp is defined as returning a "void*".
<function_type 0x4241d4fc
type <pointer_type 0x4240ea6c type <void_type 0x4240e9f8 void>
public unsigned SI
size <integer_cst 0x424074a0 constant invariant 32>
unit size <integer_cst 0x42407160 constant invariant 4>
align 32 symtab 0 alias set -1>
SI size <integer_cst 0x424074a0 32> unit size <integer_cst 0x42407160 4>
align 32 symtab 0 alias set -1
arg-types <tree_list 0x4240d150 value <pointer_type 0x4240eb54>
chain <tree_list 0x4240d138 value <pointer_type 0x4240eb54>
chain <tree_list 0x4240d120 value <integer_type 0x4240e89c SIGNED_32>>>>
pointer_to_this <pointer_type 0x4244e9f8>>



The bug is in build_common_builtin_nodes, I am testing a fix right now.

And here is the patch which gets me passed that failure but I get another
failure which I was getting before:


/Users/pinskia/src/new-fold/gcc/gcc/ada/ali.ads: In function 'ALI._ELABS':
/Users/pinskia/src/new-fold/gcc/gcc/ada/ali.ads:39: error: invariant not recomputed when ADDR_EXPR changed
&C.1716D.8324;


OK?

Thanks,
Andrew Pinski

ChangeLog:
	* tree.c (build_common_builtin_nodes): Fix the return type on
	__builtin_memcmp.

Attachment: t.diff.txt
Description: Text document


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]