Pedro Alves writes:
> [gdb list, plus a couple people]
>
> See
https://lists.gnu.org/archive/html/bug-readline/2016-04/msg00006.html
>
> (note that ml archive only refreshes every 30 mins...)
>
> On 04/14/2016 03:59 PM, Chet Ramey wrote:
> > On 4/14/16 6:47 AM, Pedro Alves wrote:
> >
> >> FYI, this is probably doing to break (at least) gdb against system
> >> readline. gdb relies on accessing a few private readline
symbols... :-/
> >>
> >> E.g.:
> >>
> >> tui/tui-io.c:437: extern int _rl_echoing_p;
> >> completer.c:1677: extern int _rl_complete_mark_directories;
> >> completer.c:1770:extern int _rl_completion_prefix_display_length;
> >> completer.c:1771:extern int _rl_print_completions_horizontally;
> >>
> >> Unless/until someone fixes these and adds the missing public APIs
> >> to readline,
> >
> > These aren't exactly the result of `missing public APIs'.
echoing_p is
> > a saved value indicating whether or not the terminal flags readline
> > inherits include ECHO -- something an application can easily check
for
> > itself.
>
> It's not that GDB wants to do something with the value. It's that
> when gdb enables/disables its TUI/curses mode (c-a x), it needs to
> save/restore that variable, here:
>
>
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=blob;f=gdb/tui/tui-io.c;h=18c648c08fdbe9a8f7d04c406c6a99e9149ca295;hb=2d059a33d890c017c8105b102a6b56ccbd6128b2#l426
>
> That's quite old code. This gdb commit:
>
> commit cc88a640ca1d0356c5feb40bb48869bab5a2bdce
> Author: Jan Kratochvil <jan.kratochvil@redhat.com>
> AuthorDate: Wed May 11 23:38:44 2011 +0000
>
> Imported readline 6.2, and upstream patch 001.
>
> (That's gdb's built-in fallback readline copy, which is
> discouraged in favor of the system one.)
>
> Did:
>
> void
> tui_setup_io (int mode)
> {
> - extern int readline_echoing_p;
> -
> + extern int _rl_echoing_p;
> +
>
>
> So looks like at some point readline_echoing_p was renamed
> to _rl_echoing_p?
>
> And that readline_echoing_p referencing code was added back in 2001
...:
>
> a198b876 gdb/tui/tuiIO.c (Stephane Carrez 2001-07-21 19:56:54
+0000 506) void
> a198b876 gdb/tui/tuiIO.c (Stephane Carrez 2001-07-21 19:56:54
+0000 507) tui_setup_io (int mode)
> a198b876 gdb/tui/tuiIO.c (Stephane Carrez 2001-07-21 19:56:54
+0000 508) {
> a198b876 gdb/tui/tuiIO.c (Stephane Carrez 2001-07-21 19:56:54
+0000 509) extern int readline_echoing_p;
> a198b876 gdb/tui/tuiIO.c (Stephane Carrez 2001-07-21 19:56:54
+0000 510)
>
> I haven't tried to remove that and see what breaks.
>
> > The others are values for settable variables and should be retrieved
> > using rl_variable_value (variable_name). Since it returns a string,
> > you can use the public macro RL_BOOLEAN_VARIABLE_VALUE to turn it
into
> > an int.
>
> The GDB commit that added those was:
>
> commit 82083d6dbbc0b2f6a76095582c6e7ffb3e06432a
> Author: Doug Evans <xdje42@gmail.com>
> AuthorDate: Sat Jan 31 14:11:54 2015 -0800
>
> Unify CLI/TUI interface to readline tab completion.
>
> This copies a lot of code from readline, but this is temporary.
> Readline currently doesn't export what we need.
> The plan is to have something that has been working for awhile,
> and then we'll have a complete story to present to the readline
> maintainers.
>
> This is again about the curses mode (TUI).
>
> Doug, could you comment?
The code in question (commit 82083d6dbbc0b2f6a76095582c6e7ffb3e06432a)
replicates the tab completion match list displayer in order to work
better with TUI. The patch includes this comment to (hopefully)
explain what's going on.
/* GDB replacement for rl_display_match_list.
Readline doesn't provide a clean interface for TUI(curses).
A hack previously used was to send readline's rl_outstream through a
pipe
and read it from the event loop. Bleah. IWBN if readline abstracted
away all the necessary bits, and this is what this code does. It
replicates the parts of readline we need and then adds an abstraction
layer, currently implemented as struct match_list_displayer, so that
both
CLI and TUI can use it. We copy all this readline code to minimize
GDB-specific mods to readline. Once this code performs as desired then
we can submit it to the readline maintainers.
N.B. A lot of the code is the way it is in order to minimize
differences
from readline's copy. */
There are two aspects to this:
1) some _rl_ state vars are referenced
2) some _rl_ functions are used
If there's a public way to access the _rl_ vars, great.
As for the _rl_ functions, I think there are two:
_rl_erase_entire_line
_rl_qsort_string_compare
Readline does export a few things like _rl_erase_entire_line,
e.g., rl_crlf, so may it'd be ok to export that one.
I think I used _rl_qsort_string_compare for simplicity.
If necessary gdb could duplicate that one I think.
That would get us past this patch.
Ultimately, there's still a lot of code duplication.
My hope is that we can reduce/remove it, but that's
perhaps a topic for another day.