Was FORTRAN buggy 4355

LarryWeiss on the Perq

My guess would be that if the Perq people had come from a unix background, the loader (and the object file format) might not have the ability to express the properties that a proper FORTRAN COMMON block requires. Certainly, these properties cannot be expressed in C data declarations.

As a old-time student of "comparative operating systems" in the 1970s, I used to marvel at how the concepts that one group of users (who had "grown up with" a particular combination of languages and systems) found simple, intuitive and obvious would appear mind-bogglingly obscure and perverse to someone from a different background.

An example: Global variables in K&R C have semantics that are equivalent to FORTRAN named COMMON, with a separate COMMON block for each variable. I.e. as far as the linker is concerned, the name space for C's global variables is not the name space for external names (such as procedure entry points) but the namespace for global CSECTs. If you have implemented FORTRAN first, the object file format may have defined this namespace with names no more than 8 characters long (or even 6) with no distinction between uppercase and lowercase. If you have implemented C first, you may have the same namespace for global labels and CSECTs, preventing a subroutine and a common block from having the same name. (I have seen both problems, though I cannot remember where. And I have never touched a Perq.)