RE: Re: union type is evil, redux

yes, i definitely have discovered compiler problems in the past. there's another message in which i wrote about this yesterday, i think. when expressions got too complicated, gcc failed to do structure assignments completely in parallel. the bugs only occur usually when you turn on error-checking and mule, because these options greatly increase the complexity of the expressions. OCaml may simply not have any code with the expression complexity that we have. ------- Original Message ------- On Fri, 20 Dec 2002 23:52:22 +0300 Dmitry Bely?wrote: "Ben Wing" <ben(a)666.com> writes: > --use-union-type Enable or disable use of a union, instead of an > int, for the fundamental Lisp_Object type; this > provides stricter type-checking. WARNING: This has > a tendency to trigger compiler bugs, especially in > combination with other features that increase the > complexity of expressions, for example `--with-mule' > and `--error-checking=all'. Crashes have been > observed with union type in combination with the > two options just mentioned under various versions of > GCC as well as MSVC++ 6.0. Furthermore, many > debuggers have problems (i.e. bugs) dealing with > unions, and even for those that don't, debugging can > be inconvenient because of no syntax for entering a > union. We recommend `--use-union-type' *ONLY* for > testing purposes, not for production builds. If you > are using `-use-union-type' and get some weird > crash, try redoing without union type. I am probably too late to jump into the discussion, but are you absolutely sure that it was _compiler_ bug but not the poorly written code whose behaviour is undefined? Do you have a small example showing such _compiler_ problem? There is a language (OCaml) those datastructures are almost entirely based on the union type, and I never heard of compiler problems related to that, although Ocaml can be build with gcc, VC++ and other C compilers under all major OSes. - Dmitry Bely