Found one more delta to unbreak build for z/os

[perry-ca]

I found we had one more delta in addition to the change in #82208 that was needed to getting the builds for z/os to work again.

[arichardson]

Thanks this looks good to me, I will merge once I've tested various configurations locally.

[arichardson]

The functions below work without int128, so this makes sense to me, but it would be good to check that this doesn't break 32-bit sparc.

[perry-ca]

@rorth can you confirm this works with 32-bit sparc (replacing your change in #101662 with this one). Thanks

[rorth]

While this does work with gcc-14, it breaks when using clang-20 instead: 26 builtins files fail with the likes of

FAILED: projects/compiler-rt/lib/builtins/CMakeFiles/clang_rt.builtins-sparc.dir/extendhftf2.c.o 
/vol/llvm/src/llvm-project/dist/compiler-rt/lib/builtins/fp_lib.h:411:2: error: Unsupported TF mode type
  411 | #error Unsupported TF mode type
      |  ^

long double on SPARC is a royal mess, unfortunately: while the SPARC psABI (both 32 and 64-bit) requires long double to be 128 bit (although no current hardware does support that) and Solaris follows the spec, Linux/sparc64 chose to ignore that, keeping long double as 64 bit. While gcc gets this right, clang never did.

The following shows the values of the relevant macros:

       	      	  	clang-20		gcc-14

  QUAD_PRECISION	defined			defined
  CRT_HAS_TF_MODE	undef			undef
  			CRT_HAS_128BIT && CRT_HAS_F128
  CRT_HAS_IEEE_TF	undef			defined
  			same as next + __LDBL_MANT_DIG__ == 113
  CRT_LDBL_128BIT	undef	       		defined
  			__LDBL_MANT_DIG__ == 113 || (__FLT_RADIX__ == 16 && __LDBL_MANT_DIG__ == 28)

  CRT_HAS_F128		undef			defined
  __LDBL_MANT_DIG__	53			113
  __FLT_RADIX__		2			2

Due to all this, dealing with long double on SPARC is fragile as hell...

[perry-ca]

Given what you have said, I would suggest the real fix for SPARC is to change the list of source files for SPARC so you exclude all of the source files related to QUAD_PRECISSION. This looks like you just need to skip adding GENERIC_TF_SOURCES to the list of source files for SPARC.

[arichardson]

So Clang and GCC disagree on the Linux ABI here? It sounds like clang should be following GCC and using 128-bit long double since that will be used for all existing Linux code?

If there is not 128-bit floating point type on sparc (at least with clang), the it sounds to me like it should not be building the tf files at all?

It looks like GCC uses IEEE 128-bit long double, so maybe we can just hoist the ifdefs above the fp_lib.h include?

[perry-ca]

@rorth Ping. Have you looked at this? I would like to get the z/os builds working again.

[rorth]

So Clang and GCC disagree on the Linux ABI here? It sounds like clang should be following GCC and using 128-bit long double since that will be used for all existing Linux code?

No: Solaris/sparc follows the ELF SPARC psABI which dictates that long double be 128-bit. While Linux/sparc, generally following the psABI, it chose to ignore that particular part of the spec and went for 64-bit long double instead. clang should indeed match the psABI/GCC/Solaris libc, but fixing this is way beyond my abilities, and LLVM SPARC maintenance these days is limited.

If there is not 128-bit floating point type on sparc (at least with clang), the it sounds to me like it should not be building the tf files at all?

There is on Solaris, but it's soft-float only.

It looks like GCC uses IEEE 128-bit long double, so maybe we can just hoist the ifdefs above the fp_lib.h include?

I've meanwhile found that the builtins situation is even messier than I thought: until LLVM 17, libclang_rt.builtins-sparc.a did contain __divtc3 and__multc3. Sometime before LLVM 18, those definitions got lost (and apparently there are no checks that the builtins interface remains stable). So instead of cementing that regression, the definitions should be restored, not removed for good. I've started looking into when the removal happened, but unfortunately so many intermediate revisions don't even build that this is very hard to do ;-(

What I believe should happen first is identify the revision that caused this breakage, than look into fixing that to continue working on SPARC together with whatever it was meant to achieve.

[perry-ca]

Thanks for the update.

Also look at the compiler side. Can the compiler generate calls to __divtc3 and __multc3 or any of the TF functions? If the compiler can't generate a call to these TF functions then it is ok to remove them from the archive even though LLVM 17 had them in it. I'd also look at what code would be in the __divtc3 in LLVM 17 if you can't provide that logfb function. How did that file compile in the first place?

When making your patch can you start with this change so we are building on top of the same code. And when you are ready ping me and I will try your change on z/OS.

[arichardson]

This looks good to me if @rorth is fine with it.

[perry-ca]

@rorth I've updated this PR with the changes I think we need for Sparc 32-bit. Can you try it.

[perry-ca]

@rorth ping. Curious if you tried this out yet?

[rorth]

I've been quite busy with all sorts of issues, including several broken build.

I've now tried this patch on sparcv9-sun-solaris2.11, first in a 1-stage build with gcc-14, then a 2-stage build also starting from gcc-14. The results weren't good: in both cases I get 3 regressions:

  Builtins-sparc-sunos :: compiler_rt_fmaxl_test.c
  Builtins-sparc-sunos :: compiler_rt_logbl_test.c
  Builtins-sparc-sunos :: compiler_rt_scalbnl_test.c

The tests fail in the same way:

In file included from compiler-rt/test/builtins/Unit/compiler_rt_logbl_test.c:4:
compiler-rt/lib/builtins/fp_lib.h:402:2: error: Unsupported TF mode type
  402 | #error Unsupported TF mode type
      |  ^

Even worse, I happened to try a amd64-pc-freebsd14.0 build off the same tree: there the build was broken:

AILED: projects/compiler-rt/lib/builtins/CMakeFiles/clang_rt.builtins-i386.dir/extendxftf2.c.o
[...]
In file included from /vol/llvm/src/llvm-project/dist/compiler-rt/lib/builtins/extendxftf2.c:13:
compiler-rt/lib/builtins/fp_lib.h:402:2: error:
Unsupported TF mode type
  402 | #error Unsupported TF mode type
      |  ^

This is getting more and more of a nightmare: it feels like completely flying in the dark, (sort of) fixing one port while simultaneously breaking at least two others. This cannot continue like this without a clear understanding how this is all supposed to work. It's just an enormous waste of time.

[perry-ca]

Thanks for trying this out @rorth. I agree with you. The builtins library touches a lot of platform uniqueness and as a result some of these problems can't be avoided. One thing I have noticed is that you can't run the unit tests for the builtins library unless you have at least one of the other higher level runtime libraries. That will be a major headache to fix.

The first three look to me like they should be excluded since these aren't being included in builtins. My quick thought is they need a line like // REQUIRES: librt_has_XXX. I see a couple other tests that define QUAD_PRECISION and have a requires line.

@arichardson any thoughts on the last one? I'm thinking we tweak the fb_libs.h a little more and use this conditional include structure for this block of code:

#elif defined(QUAD_PRECISION)
  #if defined(CRT_HAS_TF_MODE)
    #if defined(CRT_HAS_IEEE_TF)
    #else
      #error Unsupported TF mode type
    #endif
  #elif defined(CRT_LDBL_128BIT)
  #endif
#endif // *_PRECISION

I'll put this change up shortly once I try it out.

[arichardson]

This looks correct to me, but I have not tested it!

[perry-ca]

@rorth When you get a chance can you try this latest. Thanks