Kirill Ignatiev

I should add that this is a reduced test case from an issue I found in my code where a for loop run with `@threads` would use my cpu to...

I looked into it a little bit more with bpftrace, lldb, and perf. It seems that there is a loop in partr.jl in `multiq_deletemin` that assumes that the task it...

I also got this issue, here is a stack trace with sqlitebrowser-debuginfo installed: ``` Process 1171968 launched: '/usr/bin/sqlitebrowser' (x86_64) /usr/include/c++/11/bits/basic_string.h:1112: std::__cxx11::basic_string::const_reference std::__cxx11::basic_string::front() const [with _CharT = char; _Traits = std::char_traits;...

This doesn't really fix the issue because there is no backtrace saved with the exception so it's a little useless the way this pr is.

Sorry I forgot about this PR. I think it works, I've been using my own local NLopt and it throws exceptions with backtraces correctly.

I also added saving backtraces to be shown with the exception. Obviously I can't rethrow the same exception, you have to save it from the function being called by C,...

Quote from readme: > You can do this by throwing any exception inside your objective/constraint functions: the optimization will be halted gracefully, and the same exception will be thrown to...

For reference, this is what glibc seems to be doing, if I translated the code right (in works in my tests): ``` function get_FE_INEXACT() # https://github.com/JuliaLang/julia/blob/2e6d908cea8334e875ea4d22d771a6eab0eef8e7/src/jlapi.c#L413 u = zeros(Cint, 9)...

One idea I had is that the `result - xy == z` is not actually a correct way to test whether the sum `xy + z` was computed exactly, and...

@simonbyrne Right, sorry, I wasn't thinking about licenses when I posted that, I was just curious what the right approach was. They both seem to be based on fma emulation...