Request for review 7198519: Broken build, hotspot-rt win USE_PRECOMPILED_HEADER=0

[David Holmes]

On 4/10/2012 10:57 PM, harold seigel wrote:
> Hi David,
>
> Thanks for reviewing my proposed change.
>
> 1. I don't know why the other header file #include statements are
> commented out only for Windows. I suspect that the Windows build worked
> with these lines commented out. So, the comments were never removed.
>
> 2. Even though all platforms appear to need <sys/stat.h>, it is still
> potentially an operating system dependent file. Hence, its #include
> statements will still be in operating system dependent files.

FYI <sys/stat.h> is already directly included into system independent 
files eg share/vm/classfile/classLoader.hpp

Cheers,
David

> Thanks, Harold
>
> On 10/3/2012 8:27 PM, David Holmes wrote:
>> Hi Harold,
>>
>> On 4/10/2012 5:43 AM, harold seigel wrote:
>>> Summary: Broken build, hotspot-rt win USE_PRECOMPILED_HEADER=0
>>>
>>> Also deleted include statements that were commented out.
>>>
>>>
>>> open webrev athttp://cr.openjdk.java.net/~coleenp/BUG7198519
>>>
>>> bug link athttp://bugs.sun.com/view_bug.do?bug_id=7198519
>>>
>>> Tested by successfully building on Windows with USE_PRECOMPILED_HEADER
>>> turned off. Also, ran tests with build and tested with JPRT.
>>
>> This makes jvm_windows.h consistent with the other platforms regarding
>> <sys/stat.h> and fixes the problem, but it raises a couple of
>> questions for me:
>>
>> 1. Why were the other I/O header file includes commented out only for
>> windows?
>> 2. As all platforms require <sys/stat.h> Why do we not simply
>> uncomment the the include in share/vm/prims/jvm.h
>>
>> // HotSpot integration note:
>> //
>> // This file and jvm.h used with the JDK are identical,
>> // except for the three includes removed below
>>
>> // #include <sys/stat.h>
>> // #include "jni.h"
>> // #include "jvm_md.h"
>>
>> and then delete it from the platform-specific headers?
>>
>> Cheers,
>> David