RFR(XS): 8046611: Build errors with gcc on sparc/fastdebug

[Mikael Vidstedt]

On 2014-06-11 18:36, David Holmes wrote:
> Hi Mikael,
>
> On 12/06/2014 5:47 AM, Mikael Vidstedt wrote:
>>
>> Please review the below small change which fixes a couple of build
>> errors when building the sparc files with gcc.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8046611
>> Webrev: 
>> http://cr.openjdk.java.net/~mikael/webrevs/8046611/webrev.00/webrev
>>
>> Without the fix the following two errors are seen:
>>
>> hotspot/src/cpu/sparc/vm/frame_sparc.cpp:442: error: ‘frame
>> nth_sender(int)’ defined but not used
>> hotspot/src/share/vm/runtime/safepoint.cpp:771: error: ‘void
>> print_me(intptr_t*, intptr_t*, bool*)’ defined but not used
>>
>> The functions in question are debug functoins declared static. Assuming
>> we'll want to keep them around the fix is to simply remove the static
>> keyword from them.
>
> This seems so wrong. Changing them to non-static simply stops the 
> compiler from being able to deduce whether they are used or not. But 
> can they actually be used if declared non-static?
>
> Can we not tag them some way to tell the compiler to ignore them?
>
> Debugging functions can be useful even if rarely used. But no one will 
> know they exist if they aren't in the code so I'm wary of simply 
> deleting everything of that nature.

I don't know of any way to reliably tag functions as used apart from 
removing the static keyword.

I totally redid the safepoint.cpp debug function(s) part of the change. 
First of all they no longer need to handle the 32-bit SPARC case, so 
that part of the code goes away. That also enables removing the 
print_longs function (it will be the same as print_ptrs), and I did some 
additional clean up on top of that. Verified that it works by manually 
enabling the const bool and looked at the output.

I'm not sure what to do about the nth_sender function though, mostly 
because I don't know how people expect to use it. We could for example 
move it to utilities/debug.cpp and export it as the other debug function 
there, which is what I've done in the below webrev just so that we can 
see what it would look like. Gladly taking feedback on it though, 
especially from people actually having used the function in recent 
times. If it hasn't been used recently it really should go away...

Updated webrev:

http://cr.openjdk.java.net/~mikael/webrevs/8046611/webrev.01/webrev/

Cheers,
Mikael

>
> Thanks,
> David
>
>> Cheers,
>> Mikael
>>