[x86/Linux] Fix compile error in codeman

[seanshpark]

Fix compile error for x86/Linux

• add !TARGET_X86 to WIN64EXCEPTIONS directives
• fix "use of undeclared identifier 'NEED_TO_PORT_THIS_ONE'"
• fix "no member named 'GetUnwindInfo' in '_hpCodeHdr'"
• fix "use of undeclared identifier 'RUNTIME_FUNCTION__EndAddress'"

[RussKeldorph]

Please ignore the x86 compatjit and x86 legacy_backend jobs if they fail. They have been removed from future PRs.

[jkotas]

This should be left undefined (see my comment about changing the ABI in the other PR).

[seanshpark]

OK, thank you for point this.

[jkotas]

There is W(...) macro to do this without ifdefs.

[seanshpark]

OK, found it.

[seanshpark]

If I tried to block LazyGetFunctionEntry function(s) , it lead to no where, so just block the function body not to use GetUnwindInfo() and fix it properly after build is ready.

[jkotas]

Pretty much all defined(WIN64EXCEPTIONS) in this file should be defined(WIN64EXCEPTIONS) && !defined(_TARGET_X86_) - similar to what you have done under src/zap.

[jkotas]

Maybe we should invent a new #define for this.

[seanshpark]

invent a new #define

Good to hear :) . Codes seems not about Exception. Could you give a good new directive?

Pretty much all defined(WIN64EXCEPTIONS) in this file should be defined(WIN64EXCEPTIONS) && !defined(_TARGET_X86_)

The change spread to several files, threads.cpp, jitinterface.cpp, eedbginterfaceimpl.cpp and h, stackwalk.cpp and h and maybe so on. It is scary to change not knowing much.

@parjong has also tried fixing this compile error another way so if adding _ASSERTE(!"Fix This"); is inappropriate, I'll change this patch only leaving with Wmacro change.

[jkotas]

It is scary to change not knowing much.

You may be better off to just make it compile on Linux x86 same way as it compile on Windows x86 (ie without WIN64EXCEPTIONS defined). It should allow you to get hello world running, with much fewer changes. And then try to turn on WIN64EXCEPTIONS as isolated change that is not mixed up with the rest.

[seanshpark]

@ OK Thanks, I'll try again :)

[seanshpark]

Please wait, need to resolve with another PR by parjong.

[janvorli]

You may be better off to just make it compile on Linux x86 same way as it compile on Windows x86 (ie without WIN64EXCEPTIONS defined).

Yeah, I am starting to think the same. I have not realized before that the WIN64EXCEPTIONS is used at that many places in the source.

[seanshpark]

without WIN64EXCEPTIONS defined

Yes, It's getting huge. I'll close this PR and restart from without WIN64EXCEPTIONS