nrlgrx@censun1.uunet (NRL Paul Butterfield) (03/26/88)
I am interested in people's experiences with the GCC C compiler. In particular, how faithful is it to Kernighan and Ritchie? Are there any known serious bugs? Please send mail to me personally at uunet!censun1!nrlgrx. I will summarize the results in a future posting. Paul Butterfield
gcc@ICS.UCI.EDU (Douglas C. Schmidt) (03/28/88)
This reply is to Paul Butterfield's request for info on GCC. My mailer bauked at his return address... ------------------------------------------------------------ I've been testing the gcc compiler for several weeks now. When it works it generates terrific code, but there are several major compiler bugs I've discovered, including: 1. Fatal signal when attempting to pass structure parameters to functions. If the signal doesn't get you, then the bad assembly code generation on a ns32000 sequent will (for strutures as parameters, that is!). 2. Fatal signal when declaring function prototypes with variables of type char. 3. Fatal signal when dynamic-length automatic local variables are declared inside of functions (this is obviously a GCC addition to K&R). I'm sure other errors exist, but I haven't run across them yet. As long as one sticks to the vanilla K&R the compiler seems to work well. However, some of the newer ANSI and GNU extensions don't always perform as expected (see above). In summary, the compiler will be very nice, once it works! Luckily, the GNU people are very receptive to bug reports, so hopefully the problems will not endure! Doug Schmidt