[Bug build/29501] Check failed on stdlib/tst-strfrom while building for RISCV64 on a unmatched hardware

["fantasquex at gmail dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29501

--- Comment #5 from Letu Ren <fantasquex at gmail dot com> ---
As mentioned in 11.3 section of the latest RISC-V ISA manual 20191213, 

Except when otherwise stated, if the result of a floating-point operation is
NaN, it is the canonical NaN. The canonical NaN has a positive sign and all
significand bits clear except the MSB, a.k.a. the quiet bit. For
single-precision floating-point, this corresponds to the pattern 0x7fc00000.

And in section 6.3 of the latest version of IEEE 754 which is published in
2019,

When either an input or result is a NaN, this standard does not interpret the
sign of a NaN.

So, I think the problem is that IEEE 754 doesn't care about the sign bit of
NAN. And RISC-V ISA follows IEEE 754 and sets the default sign bit of NAN to
positive. The test case fails. Maybe we can change the test case a little bit,
since the sign bit of NAN doesn't matter?

BTW, I'm interested in "hook" you mentioned. Which part of code should I take a
look?

-- 
You are receiving this mail because:
You are on the CC list for the bug.