This patch is improving the alpha combine function by using int
instead of float. That function is used for example in the
switch panel to merge the transparency mask.
The change is practically indistinguishable to the human eye
for a single-pass blend but the performance gained is huge.
I've been doing some benchmark of wrlib and even implemented AVX2 support.
But the gain compared to the complexity of AVX2 is not worth,
while having int usage in that specific function is a really good trade-off.
Here the result:
Alpha Blending Performance Test
Image size: 1024x768 (786432 pixels)
Iterations: 100
AVX2 support: YES
=== RGBA Source Test ===
Original (float): 2.540 ms/frame (393.8 FPS)
Optimized (int): 1.983 ms/frame (504.2 FPS) [1.3x speedup]
AVX2 optimized: 1.843 ms/frame (542.6 FPS) [1.4x speedup]
By using int, the alpha blending in that use case is 28% faster.
This patch is implementing a new default Catmull-Rom filter
to resize images.
Catmull-Rom is a special case of cubic interpolation with B=0 and
C=0.5 (in the Mitchell-Netravali formulation).
It provides slighlty sharper results than Mitchell with the same
performance.
Catmull-Rom is a better choice for a window manager as it prioritizes
sharpness, making small elements feel crisp.
This patch is fixing some issues in how right and center aligned
wtextfields are handled.
-text selection with mouse was not working properly especially
setting and identifying the cursor position
-middle button paste was only working for left aligned text
xterm is not working properly (it's not advertising its internal icon)
if the window manager's name contains a space, seems to be specific
to xterm as xeyes and xpaint are working fine.
This patch is allowing to control the wpopupbutton entries
via keyboard up/down arrows.
It happens to me a few weeks ago during development,
wmaker crashed and I lost the mouse control.
I was stuck on that dialog box without a way to select another entry.
According to EWMH specification, the active window manager is supposed
to set some information. Those can be gathered for example with
'wmctrl -m'.
Before the patch:
$ wmctrl -m
Name: N/A
Class: N/A
PID: N/A
Window manager's "showing the desktop" mode: OFF
After the patch:
$ wmctrl -m
Name: Window Maker 0.96.0
Class: wmaker
PID: 6866
Window manager's "showing the desktop" mode: OFF
This patch adds an option to enable native CPU optimizations
by adding -march=native to the compiler flags, tuning the
generated code for the build machine at the expense of portability.
In some tests, especially on wrlib I saw 8% perf improvement.
Should be used by developers or those who recompile wmaker for their own usage.
This patch is fixing the issue reported
at https://github.com/window-maker/wmaker/issues/62
where wmaker is crashing if the font specified in WMGLOBAL is not found.
Now by default it will failsafe to DEFAULT_FONT.
The cursor is moved using right/left arrows but as it's blinking
and a XOR function is used to hide/unhide it, it happens the cursor
can be not hidden properly from its previous position.
Better to refresh the view to avoid such issues.
In case the cursor is positioned out of the textfield view
after a delete and more chars are entered, wmaker process
will reach 100% and become unresponsive.
How to reproduce:
in the run command window enter an overly long text (longer than
the current input field view). Then, press home to go back to the
beginning of the string. Then, shift-End to select all the text,
then del to delete all the text. At that point the cursor is still
out of the view and if you enter more text wmaker process will be stuck.
According to the ICCCM, a reply to a TARGETS request must be a list of atoms.
Took the chance to also fix variable naming consistency between wtext and
wtextfield which are using the same kind of code.
This patch adds optional support for compressed files and a new --ignore-unknown option
to ignore unknown image format. It also adds local filename drag-and-drop support.
And a copy current image to clipboard feature with ctrl+c shortcut.
It also fixes:
- fullscreen issue on multi monitors setup by using randr
- app icon advertised via _NET_WM_ICON
- fix UTF-8 filename usage in window title
The patch is fixing a UTF-8 truncation issue that could happen with the window title
when it was shrinked to be displayed in the window list (F11), leading to
an infinite loop.
Issue was reported at https://github.com/window-maker/wmaker/issues/61
Window Maker does not call the RegisterSession() method on GDM's D-Bus
interface, causing that GDM doesn't know that the login was successful,
which leads to problems. If X-GDM-SessionRegisters is not specified or
false, GDM registers the session itself.
I noticed this when I logged out from Window Maker, and tried to log
in into another session with GDM, it doesn't work, because the login
screen hangs as the previous session was not entered into registered
state within GDM.
If X-GDM-SessionRegisters=true is specified, GDM expects that the
session will be registered via D-Bus:
1c061b84ff
E.g. GNOME Shell calls RegisterSession() method after login:
0e37cd2ec9
Older versions behave the same way as would be X-GDM-SessionRegisters
not set or false.
Currently X-GDM-SessionRegisters=true is used only by GNOME Shell. All
other sessions omit this property, because they don't call the
RegisterSession() method on GDM's D-Bus interface after login.
Mentioned on the WMaker user mailing list, option disabled by default.
https://groups.google.com/g/wmaker-user/c/pR8P-ZYCDFo/m/Wo42U_xqBgAJ
So basically the patch is adding a new expert option in WPrefs (disabled by
default) to allow the switch panel to cycle over all the windows from all
workspaces. I believe it's useful if you are using a lot of fullscreen apps
each on different workspaces.
This reverts commit 3579c85af1.
As pointed out by David Maciejak himself, this patch triggers
an odd behavior:
"Now I cannot do a rectangular selection on the desktop with the left
click of the mouse like I used to do.
Seems the XRRQueryVersion call to get randr version is messing up with the
X events. I tried to move up the call in src/startup.c and src/main.c
instead to the point where the bug cannot be reproduced if I am putting the
XRRQueryVersion code just before the call to wDefaultsInitDomain
"WMWindowAttributes" (in src/startup.c) which is really weird."
When pango is enabled and a window title contains UTF-8 chars,
there is a chance the miniaturized icon title will cut in 2 an UTF-8 char
leading to a glitch in the title.
As mentioned on the WMaker user mailing list some time ago
https://groups.google.com/g/wmaker-user/c/95M_pb_Qlbs/m/6qJLJSqoAwAJ
The Ignore client supplied icon from the windows attributes is not working.
That's especially visible with firefox and thunderbird when they are using
NET_WM_ICON to push the embedded icon.
That patch is making sure to ignore the embedded icon if the user defined one.
As reported in bug https://github.com/window-maker/wmaker/issues/50
X11 XAllocColor() call from wGetColorForColormap() in src/resources.c is returning some errors
especially seen when running GZDoom.
TrueColor display has been the dominant standard for well over a decade, meaning almost all modern X servers default to a TrueColor visual.
The default colormap is predefined and read-only, making allocation with XAllocColor() unnecessary (and meaning no need to free it too).
The patch is checking the display visual, and in case the display is truecolor, it does not allocate or free the color, just looking up for it.
Once the patch applied, GZDoom is not reporting warnings anymore.
As reported in bug https://github.com/window-maker/wmaker/issues/34
when RandR is enabled, everytime a monitor is turned on or off wmaker is restarting
(that happens either manually or automatically with DPMS).
This behavior causes issues with xscreensaver.
How to reproduce the issue:
-install xscreensaver and lock the screen
-turn off then on the monitor manually
-you can see that wmaker was restarted and behind the xscreensaver lock window,
the desktop appears, potentially leaking information
Instead of using the RandR event RRScreenChangeNotifyMask which is too generic,
the patch is using RRCrtcChangeNotifyMask defined since RandR 1.2 (released in 2007).
In the recent RandR version, events for output (hardware) changes are propagated via
RROutputChangeNotifyMask while layout changes (like position, size, rotation)
are propagated via RRCrtcChangeNotifyMask.
The patch is purposedly not listening for RROutputChangeNotifyMask,
thus wmaker is not restarting on DPMS events anymore.
Currently, in case a new monitor is added (or removed) wmaker is not discovering it anyway
even after an automatic restart.
Either, wmaker has to be exited and restarted fully or an external tool like arandr
has to be used to configure the new monitor. So after the patch functionality remains unchanged.
Currently, it seems like the script m4/wm_i18n.m4 is passing some editing
commands to sed(1), which are not strictly conforming to POSIX[1]. Namely, the
"grouping" command:
{ command; command }
needs to have either a semicolon or a newline before the closing brace:
{ command; command; }
or
{ command; command
}
according to POSIX.
On systems which don't use the lax GNU sed by default (like OpenBSD), the
current configuration and compilation goes like this:
$ autoreconf -vif
...
sed: 1: "/po$/{s,.po,,;p}": extra characters at the end of p command
sed: 1: "/po$/{s,.po,,;p}": extra characters at the end of p command
... (etc)
$ ./configure --without-menu-textdomain CATALOGS=sr.mo LINGUAS=sr \
LIBS=-lintl MSGFMT=msgfmt --mandir=/usr/local/man
...
Translated languages to support :
configure: WARNING: No language from $LINGUAS are supported
$ gmake && doas gmake install
... (no .mo files are generated nor installed)
and so on, since the editing commands in question are affecting the processing
of .po files.
This patch proposes inserting semicolons before the closing brace in the
mentioned editing commands passed to sed(1).
[1]: https://pubs.opengroup.org/onlinepubs/9799919799/utilities/sed.html#tag_20_109_13_03
Signed-off-by: Страхиња Радић <sr@strahinja.org>
* "unknow" -> "unknown"
* "occured" -> "occurred"
Remove some entries from PO files where these entries contain spelling errors
and there are other entries that are identical except for these mistakes.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
`AC_PROG_GCC_TRADITIONAL` is obsolete and is now just an alias for `AC_PROG_CC`,
which is already defined.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
This patch adds sr.po files to Makefile.am files throughout the repository
where needed. It also adds credits for Serbian localization in README files as
needed.
Signed-off-by: Страхиња Радић <contact@strahinja.org>
Currently, menu definitions created by wmgenmenu don't include the closing
parenthesis in vgradient constructs. This patch adds the missing closing
parentheses.
Signed-off-by: Страхиња Радић <contact@strahinja.org>
This reverts commit a0b283a60f,
as it breaks saving the history in ~GNUstep/.AppInfo/WindowMaker/History
by restricting modifications to either ~GNUstep/Defaults or ~GNUstep/Library.
Thanks to Paul Selig for reporting this issue.
Since
fc63d72032 (WINGs: Fix incorrect use of macro USE_PANGO in installed header)
and
0e274dc979 (WRaster: Fix incorrect use of macro USE_XSHM in installed header)
the files wrlib/wraster.h and WINGs/WINGs/WINGsP.h are generated by the compilation,
so add them to .gitignore
As reported by Andreas Metzler, the latest API change in lib WRaster caused
a compatibility issue because the internal version number was increased.
To correctly handle this situation, this patch does 2 things:
- do not discard the 2 last number in the "c:r:a" version, because we need them;
- when calculating the version for the mapfile, use the formula that is
suggested in libtool's documentation.
The purpose of the formula is that when API is changed, if a new function
is added then the version is not incremented to reflect that we are still
compatible with current binaries, it will be incremented only when there
is a break in compatibility.
autogen.sh is reporting some warnings as below
./autogen.sh 2>&1 |grep "obsolete"
libtoolize: copying file 'm4/lt~obsolete.m4'
configure.ac:115: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:115: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:134: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:134: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:135: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:135: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:146: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:146: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:146: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:146: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:373: warning: The macro `AC_LANG_C' is obsolete.
configure.ac:373: warning: The macro `AC_TRY_COMPILE' is obsolete.
configure.ac:458: warning: The macro `AC_HEADER_TIME' is obsolete.
configure.ac:681: warning: The macro `AC_TRY_LINK' is obsolete.
configure.ac:807: warning: The macro `AC_TRY_LINK' is obsolete.
As the minimum autoconf version required is v2.69,
we need to make sure to update obsolete macros as described at
https://www.gnu.org/software/autoconf/manual/autoconf-2.69/html_node/Obsolete-Macros.html
Just a cosmetic change as the variable is currently not in use.
According to the Motif Toolkit API and MwmUtil.h, the last long
variable from the Motif WM Hints struct is currenlty used
for the status.
