using glibc2.1 with compiler other than gcc: Problems

[Roland McGrath]

> This tests exactly for what we are looking for, namely, that inlining
> is available.

That is not in fact the defining characteristic, at least to be
pedantically correct.  For standardized functions, if the compiler does not
support `extern inline' or something like it (i.e. a form of inline that
allows later extern redeclarations without complaint), we should omit the
inlines altogether.  If they are `static inline' (for a compiler that
supports inline but not `extern inline'), then later valid user
redeclarations using extern will produce warnings or errors.  If they are
`inline' (no static), that generates global function definitions in every
file that includes the header, which is not only wasteful but likely to
produce multiple definition errors.