On Sun, 2002-06-02 at 21:38, Samuel Tardieu wrote: > The package System.Img_Real is used to convert floats to string. You > can look at the expanded GNAT output using: > > % gcc-3.1 -gnatds testbug2.adb
Thanks for that tip. Will try it latter. > Well, it looks like it is a i386 specific problem, as it works just > fine on Sparc. I get the same result as the one you give as correct in > your mail. I compared the ada.numerics.aux Sin function call (I am guessing this code is i386 only, as it contains assembler code), where all the action seems to task place, in the old version and the new version, and it looks identical (note: I didn't check the Reduce function, but it wouldn't surprise me if that was identical too). So, I suspect some low level, i386 only, issue with how the Ada compiler looks at the assembler code excerpts from the Ada code. Maybe there is a problem passing parameters. However, when I copied the routine to my code, it instantly stopped working at all (it turned the Asm call into a NOP), and I half suspected at the time that maybe I was doing something wrong (do I need to use a special compiler flag or something to get Asm code to work?). -- Brian May <[EMAIL PROTECTED]> -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]