The problem is that cfgexpand.c wants to track "conflicts" between each
pair of stack variables in order to decide whether they can share a
stack slot or not -- and this test case has 100000 variables, which
means 5 billion potential conflict pairs.  This overflows a size_t value
on 32-bit host machines.

Now this happens only with -O0 -fstack-protector, because if
optimization is on, the 100000 variables are eliminated early on
(nothing is ever done to them), and if optimization is off, the pass to
compute all those conflicts is normally bypassed -- except if stack
protection is enabled, in which case all stack variables always need to
be tracked.

-- 
[armel] additional testsuite failures with -mthumb (compared to -marm)
https://bugs.launchpad.net/bugs/497295
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to