Re: Valgrind reports possibly lost memory

[Thomas Dickey]

On Mon, May 30, 2022 at 03:17:47PM -0400, Steve Litt wrote:
> Thomas Dickey said on Mon, 30 May 2022 06:01:04 -0400
> 
> >On Mon, May 30, 2022 at 10:22:21AM +0200, TheJackiMonster wrote:
> >> Hi,
> >> 
> >> I'm currently working on a messaging application using ncurses and
> >> during testing for memory leaks with valgrind I noticed reports from
> >> ncurses.
> >> 
> >> To make sure there's nothing coming from my applications code, I
> >> created a very short program to test around.
> >> 
> >> test.c:
> >> 
> >>    #include <stdlib.h>
> >>    #include <curses.h>
> >>    
> >>    int main(int argc, const char** argv)
> >>    {
> >>        WINDOW *win;
> >>        win = initscr();
> >>        endwin();
> >>        return 0;  
> >
> >That's an FAQ:
> >
> >https://invisible-island.net/ncurses/ncurses.faq.html#config_leaks
> 
> The valgrind issue is a serious problem because you can't tell whether
> the leak is in ncurses or the code you wrote. Compiling a separate
> ncurses without leak means having two different ncurses', which is a
> problem given that so much Linux software contains ncurses.
> 
> It would be wonderful if a single ncurses could operate either way at
> runtime, with an environment variable or a command line argument or the
> existence of a file, or something in a configuration file, or anything
> else detectable at runtime.

sure - but curses doesn't work that way.  It keeps memory allocated so that
applications can exit curses-mode temporarily and resume, repainting the
screen just as it was.

It's not the only library doing something like that. I saw some discussion
in the past couple of weeks here on the same theme:

        Date: Mon, 16 May 2022 17:45:07 +0100
        To: austin-group-l@opengroup.org
        Subject: Re: POSIX gettext(): lifetime of returned values

-- 
Thomas E. Dickey <dickey@invisible-island.net>
https://invisible-island.net
ftp://ftp.invisible-island.net

signature.asc
Description: PGP signature