Re: crash on resize in demos/command.c

[Lucas Gonze]

The crash is in the malloc call in the ungetch call in resizeterm, so the
cause is almost certain to be two conflicting calls to malloc.  I have not
been able to find a cause of the conflicting calls in the CDK source, so I 
don't have a patch that would solve the problem generically.

A workaround is to not do the ungetch in resizeterm:

...
    if (is_term_resized(ToLines, ToCols)) {

#if USE_SIGWINCH

/* LG edit */
#if 0
        ungetch(KEY_RESIZE);    /* so application can know this */
#endif
        clearok(curscr, TRUE);  /* screen contents are unknown */
#endif

        result = resize_term(ToLines, ToCols);
    }
...

In my application code, I can find out whether a KEY_RESIZE would have 
been received by checking to see whether the results of getmaxyx() have 
changed, so the loss of the KEY_RESIZE event doesn't matter.

A generic way to package this would be to add a publically settable flag,
then check the flag before invoking ungetch.  This might be as simple as:

in CDK:
   putenv("NCURSES_NO_GETCH=true");

in resizeterm:
...
    if (is_term_resized(ToLines, ToCols)) {

#if USE_SIGWINCH
        if( !getenv("NCURSES_NO_GETCH") )
              ungetch(KEY_RESIZE);    /* so application can know this */

        clearok(curscr, TRUE);  /* screen contents are unknown */
#endif

        result = resize_term(ToLines, ToCols);
    }
...

Though this is a workaround instead of a fix, it would probably be called
for anyway, since there are probably badly behaved user apps that call
malloc in a signal handler.

Loss of the KEY_RESIZE event is unlikely to break any existing CDK
applications, since existing CDK apps don't survive a resize and hence
don't process KEY_RESIZE events.

- Lucas