Re: mvwprintw

[Bob Rossi]

On Thu, Mar 30, 2006 at 04:24:55PM -0500, Thomas Dickey wrote:
> On Thu, 30 Mar 2006, Bob Rossi wrote:
> 
> >>In the original report I didn't understand what "curropt" is.  Can you
> >>clarify that remark?
> >
> >Yes, for some strange reason, even though I can program, I still spell
> >like a second grader. I meant to say, corrupt.
> 
> ok.  I'm not sure how the corruption is visible.  Is this a problem that 
> appears at the right margin?  For instance you could be referring to 
> unexpected characters, or misplaced characters.
> 
> (I sometimes use 'script' to record the characters that are sent to the 
> terminal - a copy of the corresponding typescript file is often as good as 
> a screenshot).

Yes. Here is a screen shot. I'm typing in the bottom window.
http://cgdb.sourceforge.net/screenshots.shtml

OK, my setup is a little different than the average. I read keys from
stdin, and pass them to readline. I use readline's alternative callback
mechanism. So I pass stuff to readline, and whatever it outputs back to
me, I store in a buffer. Then, I update the screen using mvwprintw.
So basically, I'm not sure why I can pass certain data to mvwprintw and
have it not display correctly. For each charachter typed, I do redisplay
the screen.

> >I'm using putty and my "Terminal-type string" is "xterm" and my $TERM
> >can be either xterm or xterm-color.
> 
> putty's actions on the right margin aren't exactly what xterm does.
> That's the "bw", which is used in screen optimization.

Well, I'm not outputting anything but printable char's to mvwprintw.

> >Once logged in, if I open an xterm, with TERM=xterm or TERM=xterm-color,
> >I still have the problem.
> 
> >I'm using the new ncurses.
> >
> >When I build with trace like so,
> >.../ncurses-5.5/configure CFLAGS=-g --with-shared --without-cxx 
> >--without-cxx-binding --without-ada --with-trace 
> >--prefix=/home/bar/cvs/cgdb/cgdb-cvs/ncurses/target-trace
> >and then do
> >$ ldd ./cgdb/src/cgdb
> >       libm.so.6 => /lib/tls/libm.so.6 (0xb75be000)
> >       libtermcap.so.2 => /lib/libtermcap.so.2 (0xb75ba000)
> 
> hmm - why are you linking with termcap?

I use readline, and I capture the users keystrokes manually by looking
in the info database or the termcap database. This doesn't effect
ncurses because I tell readline that the terminal type is DUMB. It only
outputs normal chars.

> >       libncurses.so.5 => 
> >       /home/bar/cvs/cgdb/cgdb-cvs/ncurses/target-trace/lib/libncurses.so.5 
> > (0xb7562000)
> >       libc.so.6 => /lib/tls/libc.so.6 (0xb742a000)
> >       /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0xb75eb000)
> >       libdl.so.2 => /lib/libdl.so.2 (0xb7427000)
> >
> >My NCURSES_TRACE env var is not set. I still get the bug.
> >If I do 'export NCURSES_TRACE=8191', I still get the bug.
> >
> >If I recompile CGDB with ncurses_g (not sure if this used
> >/usr/lib/libncurses_g.a or the one in
> >/home/bar/cvs/cgdb/cgdb-cvs/ncurses/target-trace/lib/libncurses_g.a.
> >I think it used the newone because GDB reports on the source files from
> >that library.
> 
> I guess so - unless you had a .gdbinit file that set the source path.

Nope.

> >My NCURSES_TRACE env var is not set. I do not get the bug.
> >If I do 'export NCURSES_TRACE=8191', I do not get the bug.
> >
> >So I guess it's not a function of NCURSES_TRACE. But switching libraries
> >between libncurses and libncurses_g does show a difference. Any
> >suggestions?
> 
> Aside from the traces (and compiler optimization), there shouldn't be
> a difference between the two.

OK, well I don't know what to do. I'm finding a difference. You think
it's a compiler optimization bug?

> >I've been debugging ncurses, but can't figure out what's wrong.
> 
> I usually use --with-trace when making shared libraries for debugging.

Was my configure line OK? 

I've attached both a transcript. The trace file is large. Is there a 
better number I could use to give just the info needed?
The one I have is 14MB.

Bob Rossi

typescript
Description: Text document