Re: fpurge on z/OS

[Bruno Haible]

Juergen Weber wrote in 
<http://lists.gnu.org/archive/html/bug-gnulib/2009-04/msg00165.html>:
> I am a bit appalled of having to implement all those low-level non-portable
> functions to get bash running.

I can understand this feeling...

> Or what about writing an generic implementation of the cited functions, that
> would be based on Posix API only?

This is hardly possible. fpurge's job is to "discard all pending buffered I/O",
which is related to fflush, but not the same as fflush.

> Instead of having to port fpurge and friends to all platforms, couldn't bash
> be based on Posix API only, even if that might be a performance-degradation?
> 
> #if defined SLOW_POSIX_BASED_IMPLEMENTATION 
> ...
> #else
>    fpurge and friends
> #endif

Actually, the only places where bash-4.0 calls fpurge is directly after fflush
on the same stream, and only on output streams:

  fflush (stdout);
  fpurge (stdout);

Does someone understand the purpose of this code? IMO fpurge is a nop,
right after fflush (assuming that fflush succeeded).

Bruno