sampson@cod.NOSC.MIL (Charles H. Sampson) (04/19/91)
In article <60910417115206.0004238973NC2EM@mcimail.com> 0004238973@MCIMAIL.COM (Frank Pappas) writes, while comparing several PC compilers: > Meridian has really turned their compiler around, and with the >addition of the OpenAda environment, it is my choice for a low-cost >Ada compiler. It's true Meridian occasionally suffers from some >embarrassing problems. For example, as raised in a recent net >posting, if the type mark is omitted from a subtype declaration >containing an integer range constraint, Integer is assumed for the >type mark. That's to be expected when you are turning a poor product >into a good one. As a compiler implementor, one of the games I play is trying to guess what causes other's compiler bugs. (Figuring out my own is not a game; that's work!) This Meridian mistake is not what I would expect from any compiler. If the identifier following SUBTYPE is omitted, you have a simple syntax error, which should be caught by the parser long before any of the semantic routines have the opportunity to assume anything. The only explanation that makes any sense is that there is an error in the BNF that Meridian fed into their parser generator but, if this is the case, it could have been easily fixed long ago. Charlie