RE: Core dump in the latest CDK

[Thomas Dickey]

On Wed, 4 Jan 2006, Fu Steve X wrote:

> Actually the core dump was caused by a bug in my code -- I called
> destroyCDKScreen from a callback function. But It didn't bomb before.

Among other things, I fixed the memory leaks that I could find using
valgrind in the examples (there are a few that I overlooked, and noted
for another time, of course).

> I am in the process of testing my application with the new CDK. Will let you
> know more.

thanks

> Thanks.
>
> Steve Fu
>
>
> -----Original Message-----
> From: Thomas Dickey [mailto:address@hidden
> Sent: Wednesday, January 04, 2006 8:46 AM
> To: Fu Steve X
> Cc: 'address@hidden'
> Subject: RE: Core dump in the latest CDK
>
> On Wed, 4 Jan 2006, Fu Steve X wrote:
>
> > I don't use menus. Thanks!
>
> I know that I have some email describing what you're doing with TAB,
> but it would also be helpful to see how you're adapting traverse.c
> currently.
>
> > Steve Fu
> >
> >
> > -----Original Message-----
> > From: Thomas Dickey [mailto:address@hidden
> > Sent: Wednesday, January 04, 2006 8:18 AM
> > To: Fu Steve X
> > Cc: 'address@hidden'
> > Subject: RE: Core dump in the latest CDK
> >
> > On Wed, 4 Jan 2006, Fu Steve X wrote:
> >
> > > Yes, I use CDK with Solaris Curses.
> > >
> > > Now I know what happened. As discussed in the thread below, I moved the
> > call
> > > InjectObj (curobj, key) in traverseCDKScreen around in order to register
> > > callbacks for the TAB and ESC keys. That was in the previous version of
> > CDK.
> > > Now I do the similar thing in the new CDK, moving the call InjectObj
> > > (curobj, key) to the first line in traverseCDKScreen. Callbacks for TAB
> > and
> > > ESC work but that causes core dump.
> > >
> > > http://lists.gnu.org/archive/html/bug-ncurses/2005-12/msg00019.html
> > >
> > > Is there any way to get the callbacks for TAB and ESC to work correctly
> > with
> > > traverseCDKScreen?
> >
> > That's going to take some more thinking.  In the current snapshot, I was
> > looking at complaints from three different people who wanted to change
> > traverseCDKScreen in different ways.  I addressed part of this by making
> > that a simpler function which calls reusable pieces.
> >
> > So it would be "simple" to solve your problem with ESC by making a new
> > function corresponding to traverseCDKScreen() which doesn't use menus
> > (I forget if you want to use menus).
> >
> > However, TAB is still embedded in the logic of the traverseCDKOnce()
> > function. Also, traverse.c doesn't have its own key bindings.  I think
> > that's the problem we have to address.
> >
> > One thing that I have in mind to reduce some of this is to make Cdk use
> > virtual bindings (keycodes unrelated to curses) as I did in dialog. Doing
> > that would get rid of the literal KEY_TAB in the case statement, moving it
> > to a binding table (which could then be altered).
> >
> > >> Thanks!
> > > Steve Fu
> > >
> > >
> > > -----Original Message-----
> > > From: Thomas Dickey [mailto:address@hidden
> > > Sent: Tuesday, January 03, 2006 3:33 PM
> > > To: Fu Steve X
> > > Cc: 'address@hidden'
> > > Subject: Re: Core dump in the latest CDK
> > >
> > > On Tue, 3 Jan 2006, Fu Steve X wrote:
> > >
> > > > Thomas,
> > > >
> > > > I started to look at version 20051230. Some fixes works well such as the
> > > > setCDKFocusFirst function and getting rid of multiple highlighted
> widgets
> > > > when first popping up. Thanks so much!
> > > >
> > > > Please take a look at the core dump below. Thanks.
> > >
> > > Is this with Solaris curses?  There could be some place where it returns
> > > a null WINDOW pointer that ncurses does not.  A quick check with ncurses
> > > and valgrind isn't showing me a problem.
> > >
> > > The quick fix of course would be to add a null-pointer check in
> > > fixCursorPosition(), but it would also be good to fix whatever problem is
> > > producing the null-pointer check.
> > >
> > > thanks
> > >
> > > > Steve Fu
> > > >
> > > > ------------------------ cut here -----------------
> > > >
> > > > program terminated by signal SEGV (no mapping at the fault address)
> > > > 0xffffffff7e225d50: wmove+0x0020:       ldsh     [%i0 + 6], %g4
> > > > Current function is fixCursorPosition
> > > >  215      wmove(InputWindowOf(widget), ypos, xpos);
> > > > (dbx) where
> > > >
> > > >  [1] wmove(0x0, 0x1, 0x1, 0x1b, 0x232d, 0x0), at 0xffffffff7e225d50
> > > > =>[2] fixCursorPosition(widget = 0x100280670), line 215 in "scroll.c"
> > > >  [3] _injectCDKScroll(object = 0x100280670, input = 27U), line 375 in
> > > > "scroll.c"
> > > >  [4] traverseCDKOnce(screen = 0x100185950, curobj = 0x100280670, keyCode
> > =
> > > > 27, functionKey = 0, funcMenuKey = 0x10005c1c0 =
> > > > &`assn.exe`traverse.c`checkMenuKey(int keyCode, int functionKey)), line
> > > 291
> > > > in "traverse.c"
> > > >  [5] traverseCDKScreen(screen = 0x100185950), line 370 in "traverse.c"
> > >
> > > --
> > > Thomas E. Dickey
> > > http://invisible-island.net
> > > ftp://invisible-island.net
> >
> > --
> > Thomas E. Dickey
> > http://invisible-island.net
> > ftp://invisible-island.net
>
> --
> Thomas E. Dickey
> http://invisible-island.net
> ftp://invisible-island.net

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net