Re: not bug

[Thomas Dickey]

On Wed, Aug 07, 2013 at 08:10:29PM +0400, mallachiev wrote:
> To Tomas. I have a question about
> https://bugzilla.redhat.com/show_bug.cgi?id=892674#c3. Is there any
> documented conformation that i should call fflush(stdout) or

google on "mixing printf curses" finds at the moment 940,000 hits...
A lot of those are related.  It's been that way for so long, that I
can't recall the first time the question came up.  However, standards are
weak on prohibiting things - they leave them unspecified to allow growth
in unexpected areas.

"X/Open Curses, Issues 7" says of newterm's input/output file pointers:

        It is unspecified whether Curses modifies the buffering mode of these 
file pointers.

It doesn't actually say (again unspecified) what Curses _does_ with the file 
pointers,
doesn't specify how output is flushed using the stdio functions.  Rather, the 
standard
points to specific functions in the Curses interface which flush output 
generated by
Curses calls.

There are certain low-level calls which are specified to write to stdout (putp 
for instance),
and there are similar calls which write using a caller-supplied function 
(tputs).
The standard doesn't say whether the high-level curses functions use these.

The standard is very consistent, pointing to printw as the Curses equivalent of 
printf.
It nowhere (that I can find) implies that an application can write to the 
screen using
stdio while in curses-mode.  It does mention buffering in 

As a note in the initscr/newterm description, it says

        Applications should perform any required handling of the SIGINT,
        SIGQUIT, or SIGTSTP signals before calling initscr.

ncurses of course uses this restriction to provide signal handlers.

So... (as I did, in reviewing the impact of this change, and finding a bug to 
fix -
see the change in 20130112 for putp), my advice is to pare down the assumptions 
in
the test-suite when it turns out that the standard doesn't require that detail.

> shouldn't mix stdio calls with ncurses calls. How can anybody
> understand it not looking to this bag?
> We develop tests for LSB, and there is a problem with mixing stdio
> calls with ncurses calls; possible solution is call fflush(stdout),
> but should be documented reason for it.
> 
> -- 
> Kurban Mallachiev
> Linux Verification Center, ISPRAS
> web: http://linuxtesting.org
> 
> _______________________________________________
> Bug-ncurses mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/bug-ncurses

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net

signature.asc
Description: Digital signature