[libc++abi] Apply simplify scan_eh_tab to SjLj

[kaz7]

Summary

Previous "simplify scan_eh_tab" patch, https://reviews.llvm.org/D93190,
saves landingpad if and only if the target is not using SjLj exceptions.
However, the landingpad is used by SjLj exception handler also. This
patch changes to set landingpad for both exception handlers.

[kaz7]

Adding reviewers who reviewed https://reviews.llvm.org/D93190. Thanks.

[MaskRay]

Thanks! Who is still using sjlj exceptions?

[kaz7]

VE is still using SjLj. :-)

[mstorsjo]

In D108082#2945405, @MaskRay wrote:

Thanks! Who is still using sjlj exceptions?

MinGW toolchains can use sjlj; for x86 32 bit, sjlj is still the default in GCC (although many environments configure it to use dwarf instead of sjlj).

[kaz7]

@mstorsjo Is it possible to test this on mingw with SjLj?

I've tested this on VE with SjLj. This patch fixes problems in check-libcxxabi and check-libcxx caused by previous patch. I prefer few more tests.

[mstorsjo]

In D108082#2950631, @kaz7 wrote:

@mstorsjo Is it possible to test this on mingw with SjLj?

I'm testing, but it seems like SjLj for mingw targets broke somewhere between 11.0.0 and 12.0.0 - currently trying to look into whether it is a code generation bug or a bug in libcxxabi/libunwind somewhere (and if it's a bug that is fixed by this patch).

[mstorsjo]

I can confirm that this fixes sjlj unwinding in mingw configurations too, working just as well as it did in LLVM 11.0.0. Thanks!

(This would be a great thing to backport to the ongoing 13.x branch IMO.)

[kaz7]

@mstorsjo Thank you for your help.

May I get a LGTM from an owner? Thanks!

[mstorsjo]

@ldionne I think you're needed to unblock this one?

[mstorsjo]

@phosek @compnerd - Can either of you unblock and approve this patch for the libc++abi group?

[ldionne]

This fell under my radar -- I'm not as active on libc++abi as libc++.

I'll take what @mstorsjo said as a proof that this works and is solving some issues, but I'm personally not familiar with the code.

@tstellar This is fixing a real issue on MinGW, can I cherry-pick -x to release/13.x?

[kaz7]

Thank you!

[tstellar]

In D108082#2963468, @ldionne wrote:

This fell under my radar -- I'm not as active on libc++abi as libc++.

I'll take what @mstorsjo said as a proof that this works and is solving some issues, but I'm personally not familiar with the code.

@tstellar This is fixing a real issue on MinGW, can I cherry-pick -x to release/13.x?

Sure, go for it.

libcxxabi/src/cxa_personality.cpp

 #ifndef __USING_SJLJ_EXCEPTIONS__
             if (landingPad == 0)
             {
                 // No handler here
                 results.reason = _URC_CONTINUE_UNWIND;
                 return;
             }
             landingPad = (uintptr_t)lpStart + landingPad;
-            results.landingPad = landingPad;
 #else  // __USING_SJLJ_EXCEPTIONS__
             ++landingPad;
 #endif // __USING_SJLJ_EXCEPTIONS__
+            results.landingPad = landingPad;
             if (actionEntry == 0)
             {
                 // Found a cleanup
                 results.reason = actions & _UA_SEARCH_PHASE
                                      ? _URC_CONTINUE_UNWIND
                                      : _URC_HANDLER_FOUND;
                 return;
             }