Building gdb without ncursesw?

[Paul Smith]

On Mon, 2020-10-19 at 10:21 -0400, Simon Marchi wrote:
> On 2020-10-17 6:22 p.m., Paul Smith via Gdb wrote:
> > On Sat, 2020-10-17 at 14:35 -0400, Paul Smith via Gdb wrote:
> > > There's really no need (that I can see) for ncursesw since we
> > > just always uses utf8 all the time for everything.
> 
> Looking at configure.ac, the library is chosen using this line:
> 
>   AC_SEARCH_LIBS(waddstr, [ncursesw ncurses cursesX curses])
> 
> Then, looking at my config.cache, I see this line:
> 
>   ac_cv_search_waddstr=${ac_cv_search_waddstr=-lncursesw}
> 
> So I suppose that if you pass ac_cv_search_waddstr=-lncurses to
> configure, it will just use that instead of probing.

Seems a bit dodgy.  Also, there are other places where ncursesw is
checked before ncurses:

  AC_SEARCH_LIBS(tgetent, [termcap tinfow tinfo curses ncursesw ncurses])
    ...
  AC_CHECK_HEADERS(curses.h cursesX.h ncurses.h ncursesw/ncurses.h ncurses/ncurses.h ncurses/term.h)

For my system luckily I don't need tgetent.  And, I'm not 100% sure if
the HAVE_NCURSESW_* config options will be set properly in this
situation.

It would be nice if there were a configure option to either ignore
ncursesw, or else to specifically set the library and headers location.

Oh well, I will hack at it; thanks!