Mail Archives: djgpp-workers/2000/12/17/03:11:56
At 04:06 PM 12/16/00 -0500, DJ Delorie wrote:
>> OK, no problem. Internally, filelength is actually __filelength
now,
>> with a "#define filelength __filelength" in
"/include/libc/stubs.h".
>> I can put the #define into stubs.h for __lfilelength, and comment
it
>> out if there is no such animal elsewhere.
>
>The *only* time you put something in stubs.h is if an ANSI/POSIX
>function in libc needs to call a non-ANSI/POSIX function in libc, in
>which case the non-ANSI/POSIX function needs to be listed in stubs.h.
>(or, for ansi functions calling posix-only functions).
>
>There is no other reason why a function would get listed in stubs.
>
>The "stubbed" name (__foo) is *not* part of the published API.
Well, in this case we already had a POSIX function, fcntl, calling a
non-ANSI function, filelength, which is why I would then presume that
the #define for filelength was *already* in stubs.h. Since fcntl will
now be calling lfilelength, doesn't the same rationale apply for the
location of a define for lfilelength?
---------------------------------------------------------
Peter J. Farley III (pjfarley AT dorsai DOT org OR
pjfarley AT banet DOT net)
- Raw text -