RFR(XS): 8205172: 32 bit build broken

[Doerr, Martin]

Hi David,

it's not a compiler bug. The problem is that "const intptr_t OneBit     =  1;" uses intptr_t which is 32 bit on 32 bit platforms.
We can't shift it by 48. Seems like the other usages of right_n_bits only use less than 32.

I just noticed that there's more to fix:
--- a/test/hotspot/jtreg/serviceability/jvmti/HeapMonitor/libHeapMonitorTest.c  Mon Jun 18 16:00:23 2018 +0200
+++ b/test/hotspot/jtreg/serviceability/jvmti/HeapMonitor/libHeapMonitorTest.c  Tue Jun 19 08:53:26 2018 +0200
@@ -459,7 +459,7 @@
     live_object = (ObjectTrace*) malloc(sizeof(*live_object));
     live_object->frames = allocated_frames;
     live_object->frame_count = count;
-    live_object->size = size;
+    live_object->size = (size_t)size;
     live_object->thread = thread;
     live_object->object = (*jni)->NewWeakGlobalRef(jni, object);

Best regards,
Martin


-----Original Message-----
From: David Holmes [mailto:david.holmes at oracle.com] 
Sent: Dienstag, 19. Juni 2018 07:53
To: Doerr, Martin <martin.doerr at sap.com>; hotspot-dev developers (hotspot-dev at openjdk.java.net) <hotspot-dev at openjdk.java.net>; Roland Westrelin (roland.westrelin at oracle.com) <roland.westrelin at oracle.com>; JC Beyler <jcbeyler at google.com>
Subject: Re: RFR(XS): 8205172: 32 bit build broken

On 19/06/2018 12:10 AM, Doerr, Martin wrote:
> Hi,
> 
> 32 bit build is currently broken due to:
> "trap_mask": jdk/src/hotspot/share/oops/methodData.hpp(142) : warning C4293: '<<' : shift count negative or too big, undefined behavior
> "PrngModMask": jdk/src/hotspot/share/runtime/threadHeapSampler.cpp(50) : warning C4293: '<<' : shift count negative or too big, undefined behavior

     const uint64_t PrngModPower = 48;
!   const uint64_t PrngModMask = right_n_bits(PrngModPower);

So right_n_bits(48) should expand to:

(nth_bit(48) - 1)

where nth_bit is:

(((n) >= BitsPerWord) ? 0 : (OneBit << (n)))

so the compiler is complaining that (OneBit << 48) is undefined, but the 
conditional operator will not execute that path (as BitsPerWord == 32). 
That seems like a compiler bug to me. :(

David
-----

> Please review this small fix:
> http://cr.openjdk.java.net/~mdoerr/8205172_32bit_build/webrev.00/
> 
> Best regards,
> Martin
>