This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Remove use of deprecated_init_ui_hook from quit_confirm.


On 09/06/2013 01:29 PM, Andrew Burgess wrote:
> There are only two places that deprecated_init_ui_hook is set
> (to !NULL) that I can find:
> 
>  1. In gdbtk, deprecated_init_ui_hook is used to grab a copy of
>     argv0, but is then immediately set back to NULL, and

Then, was that setting back to NULL a recent change that had
that side effect and nobody noticed?

Hmm, it's been that way for 10 years:

 https://sourceware.org/cgi-bin/cvsweb.cgi/src/gdb/gdbtk/generic/gdbtk.c?rev=1.36&content-type=text/x-cvsweb-markup&cvsroot=src
 https://sourceware.org/cgi-bin/cvsweb.cgi/src/gdb/gdbtk/generic/gdbtk.c.diff?r1=1.35&r2=1.36&cvsroot=src&f=h

Before, it used to be only cleared if not displaying a GUI, like:

static void
gdbtk_init (char *argv0)
{
...
  /* If there is no DISPLAY environment variable, Tk_Init below will fail,
     causing gdb to abort.  If instead we simply return here, gdb will
     gracefully degrade to using the command line interface. */

#ifndef _WIN32
  if (getenv ("DISPLAY") == NULL)
    {
      init_ui_hook = NULL;
      return;
    }
#endif

Guess if nobody complained thus far, people are fine with the
alternative text...

>  2. In windows-nat.c, deprecated_init_ui_hook is used to solve
>     an order of initialisation problem when creating a command
>     alias, in this case deprecated_init_ui_hook is left set.
> 
> In top.c:quit_confirm we check deprecated_init_ui_hook to detect if
> there's a GUI running.  For (1) above this will not kick in, but


> for (2) it does.... however... I don't see why this is a good thing,
> as I understand it the windows-nat.c code is not a GUI frontend for
> gdb, but is just "running-gdb-on-windows-hosts".  I find it hard to
> believe that the shorter, less informative, quit message is really
> desired... but maybe I've missed something.

Right.  That very much just looks like an unintentional side effect...

> The following patch removes the use of deprecated_init_ui_hook
> from quit_confirm, the only change I expect from this is that the
> quit message on windows hosts will fall into line with other hosts.
> 
> OK to apply?
> 

Looks fine to me.

-- 
Pedro Alves


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]