{ %opt=-CE } { %skiptarget=go32v2 } { This test generates trouble in ntvdm.exe under Windows for go32v2 executables } { Source provided for Free Pascal Bug Report 3157 } { Submitted by "Michalis Kamburelis" on 2004-06-11 } { e-mail: michalis@camelot.homedns.org } {$mode objfpc} uses SysUtils, Math; var c:Single; temp_float:double; temp_int:Integer; notcaught: integer; begin notcaught := 2; SetExceptionMask(GetExceptionMask - [exOverflow,exUnderflow,exPrecision]); try { cosh(800) =~ 1.36E+0347, this will fit in Extended but will not fit in Single or Double. So instruction below should raise Floating point overflow. But it does not (yet). } c:=cosh(800); except on E:Exception do begin Writeln('Line "c:=..." raised ' +E.ClassName+ ': ' +E.Message); dec(notcaught); end; end; temp_float:=9/200; try { The following expression triggers 'lost precision' condition; RTL has no dedicated exception class for it and maps to EInvalidOp. Note: if this will be changed to "Round(9/200)" then this whole program will run with no exception (I guess that it's because "Round(9/200)" will be calculated at compile-time). } temp_int:=Round(temp_float); except on E:Exception do begin Writeln('Line "temp_int:=..." raised ' +E.ClassName+ ': ' +E.Message); dec(notcaught); end; end; halt(notcaught); end.