Build fails on non-x86_64 configurations due to ADLC misconfig

[Erik Österlund]

Looks good.

/Erik

On 2018-05-29 13:50, Per Liden wrote:
> A few adjustments were also needed, else it didn't build with 
> pre-compiler headers enabled.
>
> http://cr.openjdk.java.net/~pliden/zgc/fix_adlc_include_zgc/webrev.0
>
> /Per
>
> On 05/28/2018 03:09 PM, Per Liden wrote:
>> On 05/28/2018 02:40 PM, Per Liden wrote:
>>> Hi Aleksey,
>>>
>>> On 05/28/2018 02:23 PM, Aleksey Shipilev wrote:
>>>> Hey,
>>>>
>>>> Seems like latest integration of C2 barriers broke the non-x86 
>>>> builds with:
>>>>
>>>> In file included from ad_aarch64.cpp:37:0:
>>>> /pool/buildbot/slaves/sobornost/zgc/build/src/hotspot/share/gc/z/zBarrierSetAssembler.hpp:39:47: 
>>>>
>>>> fatal error: gc/z/zBarrierSetAssembler_aarch64.hpp: No such file or 
>>>> directory
>>>>   #include CPU_HEADER(gc/z/zBarrierSetAssembler)
>>>
>>> Ah, I was just about to look into why our Sparc build broke :) Your 
>>
>> (Just to clarify, it broke when I tried to compile without the "ZGC: 
>> Solaris/Sparc support" patch).
>>
>> /Per
>>
>>> proposal seems like a a good workaround/fix. I'll take it for a spin 
>>> on Sparc.
>>>
>>> Thanks!
>>>
>>> /Per
>>>
>>>>
>>>> This apparently comes from src/hotspot/share/adlc/main.cpp:
>>>>
>>>> #if INCLUDE_ZGC
>>>>    AD.addInclude(AD._CPP_file, "gc/z/zBarrierSetAssembler.hpp");
>>>> #endif
>>>>
>>>> This does not seem to work, because adlc itself is built separately 
>>>> from the rest of JVM, and
>>>> INCLUDE_ZGC=0 is not passed there, as it would be the case for 
>>>> libjvm.so build via JVM_CFLAGS. (I
>>>> have idly wondered why do we even need BSAsm in C2-ish .ad, but it 
>>>> seems that .ad does use symbols
>>>> from that BSAsm).
>>>>
>>>> So, I think we can just move the INCLUDE_ZGC block straight to .ad 
>>>> template, and let the build
>>>> figure out the include during the libjvm.so compilation, when 
>>>> INCLUDE_ZGC is available, like this:
>>>>
>>>> diff -r d0ebe2e0cff4 -r e0a872b06721 src/hotspot/cpu/x86/x86_64.ad
>>>> --- a/src/hotspot/cpu/x86/x86_64.ad     Mon May 28 10:50:04 2018 +0200
>>>> +++ b/src/hotspot/cpu/x86/x86_64.ad     Mon May 28 14:06:01 2018 +0200
>>>> @@ -538,6 +538,12 @@
>>>>
>>>>   %}
>>>>
>>>> +source_hpp %{
>>>> +#if INCLUDE_ZGC
>>>> +#include "gc/z/zBarrierSetAssembler.hpp"
>>>> +#endif
>>>> +%}
>>>> +
>>>>   //----------SOURCE 
>>>> BLOCK-------------------------------------------------------
>>>>   // This is a block of C++ code which provides values, functions, and
>>>>   // definitions necessary in the rest of the architecture description
>>>> diff -r d0ebe2e0cff4 -r e0a872b06721 src/hotspot/share/adlc/main.cpp
>>>> --- a/src/hotspot/share/adlc/main.cpp   Mon May 28 10:50:04 2018 +0200
>>>> +++ b/src/hotspot/share/adlc/main.cpp   Mon May 28 14:06:01 2018 +0200
>>>> @@ -216,9 +216,6 @@
>>>>     AD.addInclude(AD._CPP_file, "code/nativeInst.hpp");
>>>>     AD.addInclude(AD._CPP_file, "code/vmreg.inline.hpp");
>>>>     AD.addInclude(AD._CPP_file, "gc/shared/collectedHeap.inline.hpp");
>>>> -#if INCLUDE_ZGC
>>>> -  AD.addInclude(AD._CPP_file, "gc/z/zBarrierSetAssembler.hpp");
>>>> -#endif
>>>>     AD.addInclude(AD._CPP_file, "oops/compiledICHolder.hpp");
>>>>     AD.addInclude(AD._CPP_file, "oops/markOop.hpp");
>>>>     AD.addInclude(AD._CPP_file, "oops/method.hpp");
>>>>
>>>> This builds on x86_64 and fixes the build for at least AArch64. 
>>>> Other .ad files (SPARC?) require the
>>>> same change, but I cannot test it.
>>>
>>>
>>>
>>>>
>>>> Thanks,
>>>> -Aleksey
>>>>