https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101508
Bug ID: 101508
Summary: Possible undefined behaviour in cpp program using
"unsigned" type starting from GCC 9
Product: gcc
Version: 10.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pnarsing at mathworks dot com
Target Milestone: ---
Hi Team,
I'm observing an optimization issue while compiling with GCC 10.3 .
On investigating figured that issue is reproducible in GCC 9 as well .
This can be reproduced by simple test code mentioned below :
------------------------------------------------------
using int_type = unsigned int; // works without "unsigned"
int_type constexpr start{0}; // works if >= 1
int_type constexpr stop{3}; // works if <= 2
int_type constexpr mult{2}; // works if != 2
int main()
{
int result = 0;
for(int_type aa = start; aa < stop; ++aa)
{
int_type const bb = aa * mult; // works if volatile
if (aa > bb)
{
return 99;
}
++result;
}
return result; // should be == stop
}
----------------------------------------------------------------
This code snippet returns 99 instead of 3 .
You can check the following link to see the issue reproducibility:
https://godbolt.org/z/zd79PPKfn
Modifying compiler option to -O1 and -Og avoids the issue.