RFR: 6879063: SA should use hsdis for disassembly

[Yumin Qi]

You are right, agent/make is historic make for engineers who only want 
build SA alone. Certainly they can do this by
make -f sa.make from different platform locations. But that is only for 
building sa-jdi.jar, for build libsaproc, they have to build after whole 
hotspot built. In hotspot build the building order is sa-jdi.jar, 
libjvm, libsaproc.  I think to build libsaproc, we can do
make -f saproc.make
But I never try that.

Thanks
Yumin

On 2012/8/30 0:34, Staffan Larsen wrote:
> My understanding is that the makefiles in the agent/make are not used 
> by normal build. Normal builds use make/<platform>/makefiles/sa.make. 
> It's unclear to me why the files under agent/make exist at all (except 
> for historical reasons).
>
> /Staffan
>
> On 29 aug 2012, at 19:22, Kelly O'Hair <kelly.ohair at oracle.com 
> <mailto:kelly.ohair at oracle.com>> wrote:
>
>> Makefiles buried in the src tree have a good chance of being 
>> completely ignored by the new build-infra project.
>>
>> We need some very solid control over all Makefiles used in the build 
>> process.
>> The only exceptions have been test, sample, and demo makefiles, which 
>> are not used in the jdk build process.
>>
>> So if these Makefiles are used in the normal hotspot build process, 
>> it would be much much better if they
>> were moved into the make directory with the rest of the makefiles.
>>
>> -kto
>>
>> On Aug 28, 2012, at 2:12 AM, Staffan Larsen wrote:
>>
>>> Thanks for picking up Tom's work and completing it. Anything that 
>>> removes 20k lines of code must be good :-)
>>>
>>> Is there a way we can write a jtreg test for this? Either by 
>>> debugging a live JVM or a core file? Having a test would be very 
>>> helpful, although it may be impossible because of the requirement to 
>>> download and build binutils. Any other way we can add automatic 
>>> testing for this?
>>>
>>> What platforms have you done manual testing on?
>>>
>>> I noticed that the makefile changes are missing from the bsd makefiles.
>>>
>>> Unrelated comment: we should remove the ia64 code from SA...
>>>
>>> Thanks,
>>> /Staffan
>>>
>>> On 27 aug 2012, at 23:07, Yumin Qi <yumin.qi at oracle.com 
>>> <mailto:yumin.qi at oracle.com>> wrote:
>>>
>>>> Hi, all
>>>>
>>>>   Can I have you code review of
>>>>   6879063: SA should use hsdis for disassembly
>>>>
>>>> http://cr.openjdk.java.net/~minqi/6879063 
>>>> <http://cr.openjdk.java.net/%7Eminqi/6879063/>
>>>>
>>>>   The SA has Java based disassemblers for x86 and sparc but amd64.  
>>>> Instead of porting to amd64 we should switch over to using hsdis 
>>>> for it like the JVM does.  This requires a new entry point into 
>>>> hsdis, decode_instructions_virtual, which separates the address of 
>>>> the code being disassembled from the buffer containing the code.  
>>>> The existing uses of decode_instructions have been updated to use 
>>>> the new interface and SA Disassembler has Java native methods that 
>>>> call into hsdis and call back up to Java to perform the 
>>>> disassembly. Also changed makefile for hsdis build for 
>>>> both(i386/amd64).
>>>>
>>>>   All the old disassembler logic was deleted since it's 
>>>> incompatible with the new disassembly interface. Also deleted are 
>>>> dbx based SA interface and few other dead files.
>>>>
>>>>   Tested by dumping full assembly from core files.
>>>>
>>>>   Reviewed-by:
>>>>   Contributed-by: Tom R (never)
>>>>
>>>>   Thanks
>>>>   Yumin Qi
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20120830/0b65cf62/attachment-0001.html