Patch to allow -nw to runemacs on w32

Patch to allow -nw to runemacs on w32

[Lennart Borgman]

[-- Attachment #1: Type: text/plain, Size: 209 bytes --]

In some situations you may want to be able to use the -nw option with 
runemacs.exe. All other options to emacs can be given to runemacs, but 
-nw currently gives a fatal error. The attached patch fixes this.

[-- Attachment #2: runemacs-allow-nw.patch --]
[-- Type: text/plain, Size: 2876 bytes --]

Index: runemacs.c
===================================================================
RCS file: /cvsroot/emacs/emacs/nt/runemacs.c,v
retrieving revision 1.13
diff -u -r1.13 runemacs.c
--- runemacs.c	29 Oct 2006 22:43:37 -0000	1.13
+++ runemacs.c	15 Nov 2006 13:30:18 -0000
@@ -51,11 +51,12 @@
   SECURITY_ATTRIBUTES sec_attrs;
   PROCESS_INFORMATION child;
   int wait_for_child = FALSE;
-  DWORD priority_class = NORMAL_PRIORITY_CLASS;
+  DWORD creation_flags = NORMAL_PRIORITY_CLASS;
   DWORD ret_code = 0;
   char *new_cmdline;
   char *p;
   char modname[MAX_PATH];
+  BOOL create_console = FALSE;
 
   if (!GetModuleFileName (NULL, modname, MAX_PATH))
     goto error;
@@ -113,12 +114,12 @@
 	}
       else if (strncmp (cmdline+1, "high", 4) == 0)
 	{
-	  priority_class = HIGH_PRIORITY_CLASS;
+	  creation_flags = HIGH_PRIORITY_CLASS;
 	  cmdline += 5;
 	}
       else if (strncmp (cmdline+1, "low", 3) == 0)
 	{
-	  priority_class = IDLE_PRIORITY_CLASS;
+	  creation_flags = IDLE_PRIORITY_CLASS;
 	  cmdline += 4;
 	}
       else
@@ -129,6 +130,27 @@
 
   strcat (new_cmdline, cmdline);
 
+  /* Look for -nw since it requires the console flag */
+  while (cmdline[0] == '-' || cmdline[0] == '/')
+    {
+      if (strncmp (cmdline+1, "nw", 2) == 0)
+	{
+          creation_flags = creation_flags | CREATE_NEW_CONSOLE;
+          create_console = TRUE;
+	  cmdline += 5;
+	}
+      else if (strncmp (cmdline+1, "-no-window-system", 17) == 0)
+	{
+          creation_flags = creation_flags | CREATE_NEW_CONSOLE;
+          create_console = TRUE;
+	  cmdline += 18;
+	}
+      else
+	break;
+      /* Look for next argument.  */
+      while (*++cmdline == ' ');
+    }
+
   /* Set emacs_dir variable if runemacs was in "%emacs_dir%\bin".  */
   if ((p = strrchr (modname, '\\')) && stricmp (p, "\\bin") == 0)
     {
@@ -140,18 +162,20 @@
 
   memset (&start, 0, sizeof (start));
   start.cb = sizeof (start);
-  start.dwFlags = STARTF_USESHOWWINDOW | STARTF_USECOUNTCHARS;
-  start.wShowWindow = SW_HIDE;
-  /* Ensure that we don't waste memory if the user has specified a huge
-     default screen buffer for command windows.  */
-  start.dwXCountChars = 80;
-  start.dwYCountChars = 25;
+  if (!create_console) {
+    start.dwFlags = STARTF_USESHOWWINDOW | STARTF_USECOUNTCHARS;
+    start.wShowWindow = SW_HIDE;
+    /* Ensure that we don't waste memory if the user has specified a huge
+       default screen buffer for command windows.  */
+    start.dwXCountChars = 80;
+    start.dwYCountChars = 25;
+  }
 
   sec_attrs.nLength = sizeof (sec_attrs);
   sec_attrs.lpSecurityDescriptor = NULL;
   sec_attrs.bInheritHandle = FALSE;
 
-  if (CreateProcess (NULL, new_cmdline, &sec_attrs, NULL, TRUE, priority_class,
+  if (CreateProcess (NULL, new_cmdline, &sec_attrs, NULL, TRUE, creation_flags,
 		     NULL, NULL, &start, &child))
     {
       if (wait_for_child)

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: Patch to allow -nw to runemacs on w32

[Jason Rumney]

Lennart Borgman wrote:
> In some situations you may want to be able to use the -nw option with 
> runemacs.exe. All other options to emacs can be given to runemacs, but 
> -nw currently gives a fatal error. The attached patch fixes this.

I don't see any error, OTOH, it launches emacs in a hidden console that 
you cannot get at, so isn't much use.

I don't see any reason why someone would want to use runemacs to launch 
emacs -nw, since the sole purpose of runemacs is to suppress the console 
window when launching emacs from an icon on Windows, so I don't think 
this needs fixing urgently. If we are going to fix this though, it would 
be better if emacs was run in the same console window rather than 
launching another one, as the latter may not be an option if the user is 
using telnet/ssh, or a full screen command prompt. Also we need to 
handle --no-window-system and the case where no GUI is available as well 
as -nw.

Re: Patch to allow -nw to runemacs on w32

[Lennart Borgman]

Jason Rumney wrote:
> Lennart Borgman wrote:
>> In some situations you may want to be able to use the -nw option with 
>> runemacs.exe. All other options to emacs can be given to runemacs, 
>> but -nw currently gives a fatal error. The attached patch fixes this.
>
> I don't see any error, OTOH, it launches emacs in a hidden console 
> that you cannot get at, so isn't much use.
>
> I don't see any reason why someone would want to use runemacs to 
> launch emacs -nw, since the sole purpose of runemacs is to suppress 
> the console window when launching emacs from an icon on Windows, so I 
> don't think this needs fixing urgently. If we are going to fix this 
> though, it would be better if emacs was run in the same console window 
> rather than launching another one, as the latter may not be an option 
> if the user is using telnet/ssh, or a full screen command prompt. Also 
> we need to handle --no-window-system and the case where no GUI is 
> available as well as -nw.
I get a fatal error, not just a hidden window. I think we should avoid 
the fatal error since the cost is low (just this little patch).

It sounds elegant to use the same console window if possible, but can we 
get that?

The code already handles --no-window-system.

Does runemacs run at all without a GUI system? Are situations where this 
occurs perhaps situations where it seems right to assume that the user 
is knowledgeable enough to just use emacs.exe instead?

The reason I looked into this was actually your little program to run 
emacsclient or emacs depending on if emacs server was available or not.

Re: Patch to allow -nw to runemacs on w32

[Juanma Barranquero]

On 11/15/06, Lennart Borgman <lennart.borgman.073@student.lu.se> wrote:

> I get a fatal error, not just a hidden window.

I get the same result that Jason: no error of any kind, just Emacs runs hidden.

BTW, the same happens with your patch if -nw is not the first
argument, for example:

  runemacs -xrm "Emacs.Background:red" -nw

                    /L/e/k/t/u

Re: Patch to allow -nw to runemacs on w32

[Juanma Barranquero]

On 11/15/06, Juanma Barranquero <lekktu@gmail.com> wrote:

> BTW, the same happens with your patch if -nw is not the first
> argument, for example:
>
>   runemacs -xrm "Emacs.Background:red" -nw

Which, I'd add, it's not easy to fix. You can

  strstr (command, "-nw")

for example, but you will never be sure the user isn't doing

  runemacs --eval "(setq my-nw 7)"

etc. To be sure that you're getting the right -nw you'd have to do a
more sophisticate command-line parsing (detecting quotes, arguments
with spaces, etc.). I don't think is worth the hassle. But it is your
time.

Though I'm with Jason that it is weird to "runemacs -nw".

                    /L/e/k/t/u

Re: Patch to allow -nw to runemacs on w32

[Lennart Borgman]

Juanma Barranquero wrote:
> On 11/15/06, Lennart Borgman <lennart.borgman.073@student.lu.se> wrote:
>
>> I get a fatal error, not just a hidden window.
>
> I get the same result that Jason: no error of any kind, just Emacs 
> runs hidden.

You are right. I also just get a hidden emacs.exe if I add -Q. Though I 
would not say a sometimes hidden, sometimes crashing emacs is very good ;-)

And at a second thought it seems natural that runemacs creates a new 
window - even if it is a console window. That is what runemacs is for, 
isn't it?

>
> BTW, the same happens with your patch if -nw is not the first
> argument, for example:
>
>  runemacs -xrm "Emacs.Background:red" -nw
That is by intent. I handle this argument like the other arguments that 
are handled in runemacs. Or, it is much easier doing it that way.

Re: Patch to allow -nw to runemacs on w32

[Juanma Barranquero]

On 11/15/06, Lennart Borgman <lennart.borgman.073@student.lu.se> wrote:

> You are right. I also just get a hidden emacs.exe if I add -Q.
Though I
> would not say a sometimes hidden, sometimes crashing emacs is very good ;-)

Hmm, I don't remember saying anything about -Q. runemacs does not
crash for me with -nw; it just runs hidden.

> And at a second thought it seems natural that runemacs creates a new
> window - even if it is a console window. That is what runemacs is for,
> isn't it?

No. Its only purpose is running Emacs *without* a console window. It
says it so in the comment at the very beginning:

 Simple program to start Emacs with its console window hidden.

 This program is provided purely for convenience, since most users will
 use Emacs in windowing (GUI) mode, and will not want to have an extra
 console window lying around.

> That is by intent. I handle this argument like the other arguments that
> are handled in runemacs.

The other arguments are not for Emacs. It does not know what to do
with "-high", "-low" or "-wait"; so runemacs processes the argument
and skips it. But -nw is to be understood both by runemacs (if your
patch applies) and Emacs, so processing it only if it's the first is a
no-no (or you'd have to document it).

> Or, it is much easier doing it that way.

Yeah. As H. L. Mencken said: "For every problem, there is a solution
that is simple, neat, and wrong." :)

                    /L/e/k/t/u

Re: Patch to allow -nw to runemacs on w32

[Juanma Barranquero]

On 11/15/06, Juanma Barranquero <lekktu@gmail.com> wrote:

>  This program is provided purely for convenience, since most users will
>  use Emacs in windowing (GUI) mode, and will not want to have an extra
>  console window lying around.

Now that I re-read that... Why does not Emacs hide its own console
when started in windowing mode? Why is runemacs.exe needed at all?

Hmm... I guess the answer is: it can, but there's no way to avoid the
"flash" of the console being created and then hidden.

                    /L/e/k/t/u

Re: Patch to allow -nw to runemacs on w32

[Jason Rumney]

Juanma Barranquero wrote:
> Now that I re-read that... Why does not Emacs hide its own console
> when started in windowing mode? Why is runemacs.exe needed at all?
If emacs is started from the command prompt in a console window, then 
hiding it is the wrong thing to do.

Re: Patch to allow -nw to runemacs on w32

[Lennart Borgman]

Juanma Barranquero wrote:
>> You are right. I also just get a hidden emacs.exe if I add -Q.
> Though I
>> would not say a sometimes hidden, sometimes crashing emacs is very 
>> good ;-)
>
> Hmm, I don't remember saying anything about -Q. runemacs does not
> crash for me with -nw; it just runs hidden.

No you did not. I just wanted to say it seems to be something in my 
startup files.

>
>> And at a second thought it seems natural that runemacs creates a new
>> window - even if it is a console window. That is what runemacs is for,
>> isn't it?
>
> No. Its only purpose is running Emacs *without* a console window. It
> says it so in the comment at the very beginning:

Oh, I did not read that. But what it does from a user perspective is to 
create a new visible window running Emacs. It would not be any big 
surprise if it did the same with a console window in my opinion.

> The other arguments are not for Emacs. It does not know what to do
> with "-high", "-low" or "-wait"; so runemacs processes the argument
> and skips it. But -nw is to be understood both by runemacs (if your
> patch applies) and Emacs, so processing it only if it's the first is a
> no-no (or you'd have to document it).

Yes, that is right. But it is quite normal for the options to come 
first. See for example grep:

     Usage: grep [OPTION]... PATTERN [FILE] ...

I think most users will put the options first even if emacs does not 
require it.

> Yeah. As H. L. Mencken said: "For every problem, there is a solution
> that is simple, neat, and wrong." :) 

I believe that is a surprisingly simple, net and right solution to 
something.

Re: Patch to allow -nw to runemacs on w32

[Juanma Barranquero]

On 11/15/06, Lennart Borgman <lennart.borgman.073@student.lu.se> wrote:

> No you did not. I just wanted to say it seems to be something in my
> startup files.

Ah. Either you were less than clear, or I was less than crisp.

> Oh, I did not read that. But what it does from a user perspective is to
> create a new visible window running Emacs.

Untrue. Emacs (without runemacs) *already* creates a visible window in
which it runs. It also creates an empty console window. With runemacs,
this additional windows isn't there. So, from a user's POV, runemacs
hides an annoying window.

> Yes, that is right. But it is quite normal for the options to come
> first. See for example grep:
>
>      Usage: grep [OPTION]... PATTERN [FILE] ...

Do you mean that in

  emacs --no-site-file -nw

"--no-site-file" is NOT an option?

> I think most users will put the options first even if emacs does not
> require it.

I don't think so. Not unless you document it (and even so).

> I believe that is a surprisingly simple, net and right solution to
> something.

I disagree.

                    /L/e/k/t/u

Re: Patch to allow -nw to runemacs on w32

[Eli Zaretskii]

> Date: Wed, 15 Nov 2006 14:44:25 +0100
> From: Lennart Borgman <lennart.borgman.073@student.lu.se>
> 
> In some situations you may want to be able to use the -nw option with 
> runemacs.exe.

What are those situations?  We cannot judge the importance of applying
this patch without knowing when the current behavior is an annoyance.

Re: Patch to allow -nw to runemacs on w32

[Lennart Borgman]

Eli Zaretskii wrote:
>> Date: Wed, 15 Nov 2006 14:44:25 +0100
>> From: Lennart Borgman <lennart.borgman.073@student.lu.se>
>>
>> In some situations you may want to be able to use the -nw option with 
>> runemacs.exe.
>>     
>
> What are those situations?  We cannot judge the importance of applying
> this patch without knowing when the current behavior is an annoyance.
>   

The situation where I thought the current behaviour was a problem was 
when specifying --alternate-editor to emacsclient. If the user for some 
reason (which I might not understand ;-) wants to run emacs in a 
console, and this should be started through the use of 
--alternate-editor then the simplest way (from a user perspective) to 
get this working might be to do something like

    set alternate_editor=runemacs.exe -nw

If my patch is applied of course. Otherwise the user just gets into the 
trouble we discussed before (hidden emacs or a crashing emacs).

Re: Patch to allow -nw to runemacs on w32

[Jason Rumney]

Lennart Borgman wrote:
> The situation where I thought the current behaviour was a problem was 
> when specifying --alternate-editor to emacsclient. If the user for 
> some reason (which I might not understand ;-) wants to run emacs in a 
> console, and this should be started through the use of 
> --alternate-editor then the simplest way (from a user perspective) to 
> get this working might be to do something like
>
>    set alternate_editor=runemacs.exe -nw

The simplest way is emacs -nw. Why would they use runemacs? If they are 
used to running emacs -nw, then they would have learnt quickly that 
runemacs is not suitable.

Re: Patch to allow -nw to runemacs on w32

[Lennart Borgman]

Jason Rumney wrote:
> Lennart Borgman wrote:
>> The situation where I thought the current behaviour was a problem was 
>> when specifying --alternate-editor to emacsclient. If the user for 
>> some reason (which I might not understand ;-) wants to run emacs in a 
>> console, and this should be started through the use of 
>> --alternate-editor then the simplest way (from a user perspective) to 
>> get this working might be to do something like
>>
>>    set alternate_editor=runemacs.exe -nw
>
> The simplest way is emacs -nw. Why would they use runemacs? If they 
> are used to running emacs -nw, then they would have learnt quickly 
> that runemacs is not suitable.

I do not think "emacs -nw" will work in all situations. I believe there 
should be a GUI version of emacsclient.exe. Does "emacs -nw" work if you 
use the GUI version?

Note: I added a GUI version to gnuclient.exe which I called 
gnuclientw.exe. The purpose of this was to avoid unnecessary console 
windows.

Re: Patch to allow -nw to runemacs on w32

[Eli Zaretskii]

> Date: Wed, 15 Nov 2006 23:02:33 +0100
> From: Lennart Borgman <lennart.borgman.073@student.lu.se>
> CC: Eli Zaretskii <eliz@gnu.org>,  emacs-devel@gnu.org
> 
> > The simplest way is emacs -nw. Why would they use runemacs? If they 
> > are used to running emacs -nw, then they would have learnt quickly 
> > that runemacs is not suitable.
> 
> I do not think "emacs -nw" will work in all situations.

When will it not work?