Re: new-clx/clx.f patch

[Sam S. <sd...@gn...>]

> * Dr. Werner Fink <jreare@fhfr.qr> [2005-09-13 11:46:00 +0200]:
>
>> > To make this work clisp.h should know about signals and the config.h
>> > of clx about signal.h.
>> > --- modules/clx/new-clx/clx.f
>> > +++ modules/clx/new-clx/clx.f	2005-09-12 17:44:44.000000000 +0200
>> > @@ -421,6 +421,9 @@
>> >  #  include <time.h>
>> >  # endif
>> >  #endif
>> > +#if defined(HAVE_SIGNAL_H)
>> > +# include <signal.h>
>> > +#endif
>> >  ##if WANT_XSHAPE
>> >  /* must include this before DEFMODULE so that DEFCHECKER will work */
>> >  #include <X11/extensions/shape.h>

so - why do you need to include signal.h?

>> > @@ -1609,6 +1612,8 @@
>> >   *  Chapter 2   Displays
>> >   * ----------------------------------------------------------------------- */
>> >  
>> > +int xlib_error_handler (Display*, XErrorEvent*);
>> > +int xlib_io_error_handler (Display*);
>> >  static Display *x_open_display (char* display_name, int display_number) {
>> >    Display *dpy;
>> >  
>> > @@ -1627,6 +1632,9 @@
>> >      DYNAMIC_ARRAY (cname, char, len + 5);
>> >  
>> >      begin_x_call();
>> > +    /* Now link in the error handler: */
>> > +    XSetErrorHandler (xlib_error_handler);
>> > +    XSetIOErrorHandler (xlib_io_error_handler);
>> >      if (strchr(display_name,':'))
>> >        strcpy(cname, display_name);
>> >      else
>> > @@ -1645,8 +1653,6 @@
>> >    return dpy;
>> >  }
>> >  
>> > -int xlib_error_handler (Display*, XErrorEvent*);
>> > -int xlib_io_error_handler (Display*);
>> >  DEFUN(XLIB:OPEN-DISPLAY, &rest args)
>> >  { /* (XLIB:OPEN-DISPLAY host &key :display &allow-other-keys) */
>> >    char *display_name = NULL;    /* the host to connect to */
>> > @@ -1679,12 +1685,6 @@
>> >                    { dpy = x_open_display(displayz,display_number); });
>> >    } else dpy = x_open_display(NULL,display_number);
>> >  
>> > -  /* Now link in the error handler: */
>> > -  begin_x_call();
>> > -  XSetErrorHandler (xlib_error_handler);
>> > -  XSetIOErrorHandler (xlib_io_error_handler);
>> > -  end_x_call();
>> > -
>> >  # if !defined(RELY_ON_WRITING_TO_SUBPROCESS)
>> >    disable_sigpipe();
>> >  # endif
>> 
>> you are moving XSetErrorHandler/XSetIOErrorHandler from OPEN-DISPLAY
>> to x_open_display - why?

so, why?

>> can then be called before XOpenDisplay?
> AFAIK those handler can be set before XOpenDisplay().

ok.  is there any benefit in doing so?

> If the Xlib runs only on an error if the display
> was opend with XOpenDisplay().

I don't understand this phrase. sorry.

>> 
>> > @@ -7196,6 +7205,8 @@
>> >  
>> >    pushSTACK(find_display (display));
>> >    fehler (error, "IO Error on display ~S.");
>> > +  end_callback ();
>> > +  return 0;			/* Make gcc4 happy */
>> >  }
>> >  
>> >  int xlib_after_function (Display *display)
>> 
>> fehler is declared to be non-returning.
>> I don't see any gcc 4 warnings here.
>
> Nevertheless IMHO the end_callback() should
> be called to be sure that writing_to_subprocess
> is set to false.

fehler does not return.
whatever comes after it will not be executed anyway.

>> > --- modules/clx/new-clx/config.h.in
>> > --- modules/clx/new-clx/configure
>> > --- src/unixconf.h.in
>> 
>> these are generated files, please do not send a diff for them.
>
> Only modules/clx/new-clx/configure was generated
> with autoconf and modules/clx/new-clx/config.h.in.
> The patch includes the detection of signal.h
> required for using writing_to_subprocess.

I do not understand what you are talking about.
all 3 files are generated by autoconf - just look at their headers.
why is signal.h needed for writing_to_subprocess?


-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.memri.org/> <http://truepeace.org> <http://www.iris.org.il>
<http://www.palestinefacts.org/> <http://www.savegushkatif.org>
An elephant is a mouse with an operating system.