[libc] implicit conversion error on rv32

[mikhailramalho]

Some tests are failing to build on rv32 due to:

/home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/FPBits.h:760:40: error: implicit conversion loses integer precision: 'int' to 'size_t' (aka 'unsigned int') [-Werror,-Wimplicit-int-conversion]
  760 |       result.set_significand(number >> -ep);
      |                                     ~~ ^~~
/home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/generic/FMod.h:221:19: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::internal::FPRepImpl<__llvm_libc_21_0_0_git::fputil::FPType::IEEE754_Binary128, __llvm_libc_21_0_0_git::fputil::FPBits<long double>>::make_value' requested here
  221 |       return FPB::make_value(d, e_y - 1);
      |                   ^
/home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/generic/FMod.h:286:18: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::generic::FMod<long double>::eval_internal' requested here
  286 |     FPB result = eval_internal(sx, sy);
      |                  ^
/home/mgadelha/tools/llvm-project/libc/src/math/generic/fmodl.cpp:17:46: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::generic::FMod<long double>::eval' requested here
   17 |   return fputil::generic::FMod<long double>::eval(x, y);
      |                                              ^
1 error generated.

[llvmbot]

@llvm/issue-subscribers-libc

Author: Mikhail R. Gadelha (mikhailramalho)

Some tests are failing to build on rv32 due to: ``` /home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/FPBits.h:760:40: error: implicit conversion loses integer precision: 'int' to 'size_t' (aka 'unsigned int') [-Werror,-Wimplicit-int-conversion] 760 | result.set_significand(number >> -ep); | ~~ ^~~ /home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/generic/FMod.h:221:19: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::internal::FPRepImpl<__llvm_libc_21_0_0_git::fputil::FPType::IEEE754_Binary128, __llvm_libc_21_0_0_git::fputil::FPBits<long double>>::make_value' requested here 221 | return FPB::make_value(d, e_y - 1); | ^ /home/mgadelha/tools/llvm-project/libc/src/__support/FPUtil/generic/FMod.h:286:18: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::generic::FMod<long double>::eval_internal' requested here 286 | FPB result = eval_internal(sx, sy); | ^ /home/mgadelha/tools/llvm-project/libc/src/math/generic/fmodl.cpp:17:46: note: in instantiation of member function '__llvm_libc_21_0_0_git::fputil::generic::FMod<long double>::eval' requested here 17 | return fputil::generic::FMod<long double>::eval(x, y); | ^ 1 error generated. ```

[mikhailramalho]

This fixed it for me, but I'm not sure it's the right fix:

diff --git a/libc/src/__support/FPUtil/FPBits.h b/libc/src/__support/FPUtil/FPBits.h
index bee8d0a8dc47..09aaf62bde44 100644
--- a/libc/src/__support/FPUtil/FPBits.h
+++ b/libc/src/__support/FPUtil/FPBits.h
@@ -757,7 +757,7 @@ public:
       result.set_significand(number);
       result.set_biased_exponent(static_cast<StorageType>(ep + 1));
     } else {
-      result.set_significand(number >> -ep);
+      result.set_significand(number >> -static_cast<size_t>(ep));
     }
     return RetT(result.uintval());
   }

[michaelrj-google]

I think the correct fix should have the minus sign inside the cast. To reach this line the value of ep must be negative.

[lntue]

I think the correct fix should have the minus sign inside the cast. To reach this line the value of ep must be negative.

Yes, the minus sign should be inside the cast. And probably we should use unsigned instead of size_t if it does not cause trouble.

[mikhailramalho]

static_cast(-ep) builds without issue, running the experiments now.