Erstad@HI-MULTICS.ARPA (05/22/87)
Add to the list of things Apollo didn't tell you... The Pascal compiler has had several annoying and potentially dangerous problems. Passing an integer constant to a procedure expecting a formal parameter of type real gives an incorrect answer. No errors, etc. Just the wrong answer. We came very close to shipping a bunch of bad data to a customer because of this. The data was close enough that it passed a sanity check but was wrong. Of course, part of that was our own fault for not spending more time in testing. Other compiler problems are less serious but include spurious warning messages (i.e. condition being warned about doesn't exist) and compiler failure while attempting certain optimizations for the 68020 machines. At leaset it knows it fails, instead of generating bad code. (Incidentally, all the above are based on 6.1301 and 6.1413). The shell has problems doing chpat and fpat operations involving multiple at signs. The documentation and the operation mismatch on what should / does happen. This has been around since 1984 or 1985 - still not fixed (9.2.3).