[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Marcus Daniels writes:
> Joerg-Cyril> I wonder why a parameter that seems to need local
> Joerg-Cyril> adjustement has found neither entry nor comment into
> Joerg-Cyril> config.lsp.
> I don't see any reason this parameter needs to exist at all.
> Anyone care to comment? I'd be inclined to just the UNIX code for
> everyone (localtime vs gmtime). It works for DOS, at least.
DOS and UNIX is not all of the world. You cannot assume some
apparently standard C call on every platform.
You shouldn't even assume the existence of _any_ C function on any
system, for the following reason:
When building CLISP, one (of the many) thing(s) Bruno Haible had in
mind was portability and that many compilers libraries are broken
somewhere. He didn't want to let CLISP rely on a broken C library,
behaving differently when compiled with different compilers and their
libraries or with different versions of the same compiler. He tried to
exclusively use kernel or system calls.
For instance, the Amiga version of CLISP doesn't rely on any compilers
C library. I only had to write a small startup code and all the rest
is in the CLISP source. It doesn't use any malloc, strcmp, strcpy,
strlen, sprintf, fprintf or other such common C functions. It doesn't
use any time functions other than kernel ones. And it works with any
version of the OS that most users can remember. Current ports to other
platforms work the same way, only using the C library where
unavoidable (?), using the native system I/O, memory allocation, time
> Unless someone objects, I'll just remove this parameter and rely
> on the user setting the timezone with environment variables.
Why not just let that code remain in the source and bring it to the
users attention by putting this variable into config.lsp, like I
suggested in my previous mail? Not everybody has a TZ variable, uses a
TZ variable or sees the use for a TZ variable.
I strongly object removing this working code from CLISP.
I object intrdoducing into CLISP code the unnamed assumption that
everybody has some standard (maybe even worse: ANSI) C functions. The
way CLISP works so far is to define local files (unix.d, msdos.d etc.)
where #defines are kept and to write separate code in the different
modules where it is necessary.
If several platforms can share the same code then let them do so, but
remember that others may need special case code.
No offense intended,