Gil, I agree; that sounds like a defect (or enhancement? :-)
On Mon, Jul 15, 2013 at 10:38 AM, Paul Gilmartin <[email protected]>wrote: > On Mon, 15 Jul 2013 09:23:40 -0500, Kirk Wolf wrote: > > > >Maybe that is one of the "enhancements" ? :-) > > > Exoskeletal "enhancement"? Is there any rationale for making the > ASCII behavior different from the EBCDIC? > > >Actually this seems to be an issue with the stream buffering strategy. > >See the C/C++ RTL Ref for "setvbuf()". Maybe you need to explicitly call > >setvbuf() for stdout/stderr with __LIBASCII ?? > > > In fact, "setvbuf( stdout, NULL, _IONBF, 4096 );" repairs the lazy > buffering. > Surprisingly, _IOLBF lacks that effect. Feels like a bug? > > loc. cit.: > _IOLBF > Line buffering is used for text stream I/O and terminal I/O. The > buffer is > flushed when a newline character is used (text stream), when the buffer > is full, or when input is requested (terminal). The value for size > must be > greater than 0. > > Hmmm. On a hunch, I added "...\025" to the end of each format string. > Now, the buffer gets flushed, even with _IOLBF. Ah! It's that kind of > bug. > > Guess what I hate! > > Thanks, > gil > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
