http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
Bug #: 50115
Summary: Integer test optimised away at -O2 by the VRP
Classification: Unclassified
Product: gcc
Version: 4.5.1
Status: UNCONFIRMED
Severity: major
Prior
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #1 from Balint Szente 2011-08-18
09:56:54 UTC ---
Created attachment 25048
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25048
Suse's G++ exact version and other info
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #2 from Balint Szente 2011-08-18
09:57:34 UTC ---
Created attachment 25049
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25049
Preprocessed file with Arch Linux's GCC 4.6.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #3 from Balint Szente 2011-08-18
09:58:24 UTC ---
Created attachment 25050
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25050
Arch Linux's G++ exact version and other info
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #5 from Balint Szente 2011-08-18
11:13:08 UTC ---
Yes. Is G++'s VRP considering that "value" will be always positive, and for
this reason will never be equal to ::min()?
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #7 from Balint Szente 2011-08-18
11:59:59 UTC ---
Thanks for your prompt reply. I understand that my test case is undefined
behavior from the C/C++ standard's perspective.
I disregarded that it is not allowed to make any overflows. U
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #8 from Balint Szente 2011-08-18
12:44:53 UTC ---
Just for the record, for clarifying this for me and other people... There are
two statements in C++ standard:
* Chapter 5
"If during the evaluation of an expression, the result is not
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50115
--- Comment #11 from Balint Szente 2011-08-18
13:20:18 UTC ---
Thanks for the explanation and for your time.
