On Mon, Sep 16, 2013 at 10:09:02AM +0100, Nicholas Clark wrote: > It turns out* that $Config{fflushNULL} on HP-UX is *STILL* undef, because > fflush(NULL) wrongly closes STDIN if STDIN is a pipe (which is what the > Configure test probes for). Is it worth documenting the specific failure case more clearly? Something like this? diff --git a/U/perl/fflushall.U b/U/perl/fflushall.U index 783a7f7..a4398dd 100644 --- a/U/perl/fflushall.U +++ b/U/perl/fflushall.U @@ -13,8 +13,9 @@ d_sysconf i_unistd d_stdio_stream_array stdio_stream_array i_stdlib ?MAKE: -pick add $@ %< ?S:fflushNULL: -?S: This symbol, if defined, tells that fflush(NULL) does flush -?S: all pending stdio output. +?S: This symbol, if defined, tells that fflush(NULL) correctly flushes +?S: all pending stdio output without side effects. In particular, on some +?S: platforms calling fflush(NULL) *still* corrupts STDIN if it is a pipe. ?S:. ?S:fflushall: ?S: This symbol, if defined, tells that to flush @@ -24,8 +25,9 @@ ?S: even be probed for and will be left undefined. ?S:. ?C:FFLUSH_NULL: -?C: This symbol, if defined, tells that fflush(NULL) does flush -?C: all pending stdio output. +?C: This symbol, if defined, tells that fflush(NULL) correctly flushes +?C: all pending stdio output without side effects. In particular, on some +?C: platforms calling fflush(NULL) *still* corrupts STDIN if it is a pipe. ?C:. ?C:FFLUSH_ALL: ?C: This symbol, if defined, tells that to flush Nicholas ClarkThread Previous | Thread Next