https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=277612

--- Comment #4 from Paul Floyd <[email protected]> ---
This issue resurfaced last week.

We were using "0.0/0.0" as our NaN generator. Two fun facts. GCC generates code
that does the division. On i386 and amd64 the result is a NaN with a negative
sign bit. clang does compile time constant folding and produces a positive NaN,

See this regression test failure and discussion:
https://sourceforge.net/p/valgrind/mailman/valgrind-testresults/thread/202605070526.6475Q2BE072333%40localhost.orange.fr/#msg59331691

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to