Optimizer bug when using absolute

Summary

wrong assembly is generated on linux with -O3

System Information

x86_64-linux, fpc 3.2.2

Steps to reproduce

    program project1;
     
    {$mode objfpc}{$H+}
     
    var ind:integer=0;
    const vals:array[0..1] of integer=(1052909725,1092805937);
     
    function GetLong:integer;
    begin
      Result := vals[ind];
      inc(ind);
    end;
     
    function GetDouble: double;
    type TpDoubleRec=packed record v1,v2:integer;end;
    var V:TpDoubleRec absolute Result;
    begin
      V.v1 := GetLong;
      V.v2 := GetLong;
    end;
     
    var r:double;
    begin
      r := GetDouble;
      if round(r*100000)<>61916062257 then
        writeln('wrong')
      else
        writeln('ok');
    end.

What is the current bug behavior?

Compiling on windows as i386-win32:

• without optimizing or with -O3: writes ok to console

Compiling on linux as x86_64-linux:

• without optimizing: writes ok to console
• with -O3: writes wrong to console (and r is widely off, its roughly 2.15320923002915E-317)

Full discussion including workarounds here: https://forum.lazarus.freepascal.org/index.php/topic,62766.msg474801.html