[patch #1768] [Patch #1768] Adds non-std. resolution support and ncursesw bug-fixes

[noreply]

Patch #1768 has been updated. 

Project: 
Category: None
Status: Open
Summary: Adds non-std. resolution support and ncursesw bug-fixes

Follow-Ups:

Date: Mon 07/28/2003 at 04:18
By: marco_g

Comment:
Changelog entry:

2003-07-18  Marco Gerards  <metgerards@student.han.nl>

        * console.c (cons_vcons_update): Call setres for every display
        driver before the driver is used.
        (cons_vcons_set_cursor_pos): Likewise.
        (cons_vcons_clear): Likewise.
        (cons_vcons_write): Likewise.
        * display.h (display_ops): New  interface set_resolution.       
        * vga.c (vga_disp_op): Set NULL for set_resolution.
        * ncursesw.c (ncursesw_set_resolution): New function.
        (ncursesw_displ): Add ncursesw_set_resolution.
        (ncurses_lock): Make variable static.
        (current_width): New variable.
        (current_height): Likewise.

        (conspad): New variable.
        (input_loop): Use conspad instead of (the default) stdscr.
        (mvwputsn): Likewise.
        (ncursesw_update): Likewise.
        (ncursesw_set_cursor_pos): Likewise.
        (ncursesw_scroll): Likewise.
        (ncursesw_write): Likewise.
        (ncursesw_driver_start): Likewise. Initialize conspad.

        (ncursesw_driver_start): Remove endwin call at the end of the
        function.

        (ncursesw_set_cursor_status): If the status can not be set, try
        another status that can sanely be used instead of the unavailable
        status.

        (ncursesw_write): Make it wrap around the edge.

-------------------------------------------------------

Date: Thu 08/07/2003 at 00:53
By: marcus

Comment:
Please put the changelog before the patch in the patch file, so I can copy it 
from there.

Put a cons_vcons_set_dimension callback into the console, which is called from 
libcons refresh function (and possibly if the change records indicate the 
dimensions have changed, but that is not supported yet and isn't as easy, just 
focus on refresh and one-time initialization for now).

Then get rid of all these calls to the set resolution function.

The drivers also get a set_dimension function.  Resolution is the wrong word 
here (a resolution is for example pixel per inch, while this is really a 
dimension, more text cells will require a larger screen).

Please add keys to scroll the pad around, that allows you to at least play a 
bit with larger dimensions than screensize in ncursesw.

Please revert the breaking up of line in write of ncursesw, it shouldn't be 
required with pads.

After removing this endwin() you can just as well return 0 instead ERR.  'x' 
should release the lock and call console_exit() instead endwin and exit().



-------------------------------------------------------

Date: Thu 08/07/2003 at 00:56
By: marcus

Comment:
Oh, and please imlpement the set_dimes call for the VGA driver as well, and 
make it truncate/wrap correctly for different dimensions.  We won't be able to 
implement a full pad window as easily as in ncursesw, but at least for lower 
dimensions it can be used and for larger dimensions it will at least show the 
upper left corner properly.

-------------------------------------------------------

Date: Fri 08/08/2003 at 22:05
By: marco_g

Comment:
He is the new patch that implements everything you asked for.

M-w + j,k,l,i can be used to scroll the pad. Because this sucks when doing real 
work I've added an "autoscroll" option. It scrolls the pad automaticly, 
depending on the cursor position. Autoscroll can be enabled/disabled with "C-w 
a".

For the vga driver the changes were a bit more difficult. For example scrolling 
is disabled if the vcons is bigger than the physical screen.

cons_vcons_set_dimension also deallocates the scarce resources now so 
cons_vcons_clear is not called anymore in cons_vcons_refresh. I did think about 
this, this is not a mistake. It is not possible to deallocate the resources 
before cons_vcons_set_dimension and deallocating it after the call is also 
impossible.

Please notice that this patch is actually two patches + changeslog entries.

Please look at my changelog entry, it seems a bit to big for me, possibly 
because I split it up in logical sections. (First part for set_dimensions 
related stuff, 2nd part for pad related stuff, the other parts are for 
bugfixes, etc.). I appreciate comments. :)

-------------------------------------------------------

Date: Sat 08/09/2003 at 19:22
By: marcus

Comment:
I still have a couple of issues:

* Never split a changelog entry with an empty line if you don't start a new 
entry with a file "* file.c (foo)" etc.
It doesn't make sense.  If they are truly separate issues, then repeat the "* 
file.c" part.

You have this comment:

+  /* Change the dimension of the physical screen to or one that can
+     display the vcons with the size of WIDTH * HEIGHT. If the
+     physical screen already has the right resolution do
+     nothing. Deallocate all scarce resources occupied by the screen.
+     */

Please change the wording to be more clear.  The "to or" construct is 
confusing.  Then never end a comment with "*/" on its own line, use emacs to 
properly break the lines (Meta-TAB).  And please put two spaces after a ".".  
This also will allow emacs to do proper line breaking (because it can separate 
initials and abbreviations from sentence endings).

I think the current width and height will be automatically initialized by 
set_dimension.  We can make it the protocol that it must be called once before 
any output happens.  So they don't need to be initialized to the magic values 
80 and 25.  Likewise for the VGA driver.

In the autoscroll comment, there again is a "." with only one space after it.

Don't call the function ncurses_refresh, because it is confusing (is it a 
function in ncurses?).  Call it refresh_screen.

I liked the previous visibility code better.  Checking for an error and then 
doing the right thing is better than unconditionally do something unnecessary.

In the VGA driver, if the cursor position is outside the visible area, the 
cursor must be switched off (made invisible) internally, and independent of the 
actual cursor status.  If it is put back into the visible area, it must be 
turned on again.

I see what you mean with the scrolling.  That is in fact a problem.  We have to 
fix that (either by keeping a copy of the content, or by somehow being able to 
return an error indicating that no scrolling is possible, and the data needs to 
be supplied - this will be hard to do on a per-driver basis, though).  Please 
put a FIXME in there, and the reason why we can't scroll.

The comment in the VGA set_dimension is confusing, can you reword this?  Like: 
"FIXME: Should support greater dimensions by changing the video mode." or 
something.

For clear, I think it is ok to have set_dimension clear the screen implicitely. 
 But don't say "scarce resources" in the comment, just say that this should 
include a clear on the whole (old) screen, and that it will be followed by a 
write that covers the whole (new) screen.



-------------------------------------------------------

Date: Mon 08/11/2003 at 22:22
By: marco_g

Comment:
Thanks for your comments. Here is a new patch.

I'm using the visibility code that I've used in the first patch again. I've 
also improved the vga_display_write function, the one in my first patch didn't 
work right when using a vcon that is bigger than the physical screen.


-------------------------------------------------------

-------------------------------------------------------
For more info, visit:

http://savannah.gnu.org/patch/?func=detailpatch&patch_id=1768&group_id=30

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/