RFR(s): 8221408: Windows 32bit build build errors/warnings in hotspot

[Thomas Stüfe]

On Tue, Mar 26, 2019 at 7:53 AM David Holmes <david.holmes at oracle.com>
wrote:

>
> >
> >     I have to wonder if we can simply appease the compiler by doing this:
> >
> >     address os::current_stack_pointer() {
> >           int dummy;
> >     !   intptr_t addr = (intptr_t) &dummy;
> >     !   return (address) addr;
> >     }
> >
> >     ?
> >
> >
> > I find explicitly disabling the warning much clearer to read. Plus, even
> > if we could satisfy the compiler now, this may break with the next
> > compiler update.
>
> I agree with the latter statement, but generally I prefer trying to
> convince the compiler it really has nothing to worry about. :)
>

At some point its like arguing with the cat. Then, a clear directive is
better :)


>
> >     ---
> >
> >        src/hotspot/share/classfile/classFileParser.cpp
> >
> >     Ok. As Kim mentioned this had been flagged previously but left alone
> (I
> >     thought perhaps the gcc 8.2 testing ...).
> >
> >     ---
> >
> >     src/hotspot/share/runtime/vmStructs.hpp
> >
> >     Adding the cast is at least consistent with the int32_t casts we add
> >     for
> >     GENERATE_VM_INT_CONSTANT* but it does make me think we should then
> also
> >     add it to GENERATE_PREPROCESSOR_VM_LONG_CONSTANT_ENTRY.
> >
> >
> > Okay, I can do that. Lets see what Kim thinks about the whole thing.
>
> Yeah. I'm still very curious as to why win-64 doesn't include those
> additional enum values.
>
> The error seems to indicate that it thinks the anonymous enum is larger
> than uint64_t - which seems surprising to say the least. :)
>
>
Yes. Frustrating. I'm out of answers and believe this is a fluke in the
compiler.

Cheers, Thomas


> Cheers,
> David
>
>