https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70018
--- Comment #6 from Sanjoy Das ---
(In reply to Andrew Pinski from comment #5)
> Oh I see pure/const behavior.
>
> The problem is more complex, in that in one TU, the comdat function is
> figured out to be pure/const so we remove the store befor
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70018
--- Comment #4 from Sanjoy Das ---
(In reply to Andrew Pinski from comment #2)
> So in summary what you are seeing is two effects going into effect here:
> undefined behavior of division by 0 and ODR.
There is no division by zero (or any other
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70018
--- Comment #3 from Sanjoy Das ---
(In reply to Andrew Pinski from comment #1)
> In C++ code, the one definition rule says that all TU that contains an
> inline function, they need to have the same definition. If they have
> different definition
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: sanjoy at playingwithpointers dot com
Target Milestone: ---
I don't grok gcc internals so I cannot write a terribly well-informed
bug report, but GCC 5.3.0 seems to miscompile
https://github.com/sanjoy/comdat-ipo
