Comments (14)
The "default" XFT rendering with Motif on MacOS is slow, but only for drawing labels. The rest of the software speed is normal. So it is not that dramatic. But opening menus is not snappy, and scrolling long lists is really slow. I think your program will not be terribly affected with XFT slowness, even if you use unpatched Motif. Try it for yourself and see.
As for the fact that my patch was not integrated and nobody bothered to reply or check it - it just shows that active Motif development has stopped. I have bugged ICS from time to time with pleas to fix some obvious bugs, which were reported with solution in bugzilla. A couple of times they did it, but now they don't even remove obvious spam from Bugzilla, which renders it almost useless. Bugzilla was also down for months and nobody seemed to bother.
from classic-colors.
@justinmeiners thank you Justin, for pushing for inclusion of this bug fix.
I will just point to the parts of ISC answers which are not correct in my opinion.
The wording should be:
"The bug is reported to be relevant to macOS or remote X connection only."
So it is still quite relevant, since Motif is one of the old toolkits, which still work perfectly over network, and people do use it that way. A lot of older scientific software is still Motif based, and MacOS usage in academia is high, now that classical Unix workstations are thing of the past.
RedHat has created a similar patch AND IT IS ALREADY INCLUDED in the latest Motif. Problem is, that it was not applied to all the cases, so it shows as slow performance on XmList widget with XFT fonts. Hence slow opening of Menus, or slow scrolling of textual lists (on MacOS or remote connection).
The bug consists in the implementation of drawing of XmLabel, which queries background color from server, in order to draw antialiased text with XFT (for blending foreground and background colors during antialiasing). This query causes round trip from X Server, and this incurs the speed penalty (on MacOS with XQuartz or on remote connection). On Linux, it is not noticeable, because Xlib is more optimized. But it is still a bug, since any application, which floods the X Server with Xlib requests, which need round trip, are poorly written applications, with bad understanding of X11. To quote Wikipedia on X11:
"X's design requires the clients and server to operate separately, and device independence and the separation of client and server incur overhead. Most of the overhead comes from network round-trip delay time between client and server (latency) rather than from the protocol itself: the best solutions to performance issues depend on efficient application design."
And if library (Motif) does this mistake, then even carefully written Motif applications will have poor performance.
The corrections consists in simply caching the background color from last drawing of label (thus avoiding XQueryColor and round-trip), so repetitive drawing of XFT text on the same background color (which is the most typical case) becomes fast.
from classic-colors.
This is how the program looks with XFT fonts.
from classic-colors.
I will have a PR up soon that I could use your feedback on.
Here are some resources for myself which helped me understand what this is:
http://www.motifdeveloper.com/tips/tip2.html
https://sourceforge.net/p/cdesktopenv/wiki/FontsWithXFT/
https://docs.prolifics.com/panther/html/cfg_html/motifdef.htm
https://wiki.gentoo.org/wiki/Xft_support_for_GNU_Emacs
https://motif.ics.com/forum/developers/utf8-support
I'll also add the Motif XFT PDF since it's only available on internet archive (and small):
from classic-colors.
This is a great suggestion. What do you think about the solution in PR #14 ?
from classic-colors.
I think the solution is fine.
As XFT was added to Motif in the sunset of its active development, the documentation is very scarce. You have collected the right bits of documentation.
One note for Mac users - Motif on Mac works slowly with XFT, due to some interaction of XQuartz and core MacOS drawing subsystem. If you apply this patch to Motif, program with XFT will work at decent speed.
http://bugs.motifzone.com/show_bug.cgi?id=1715
from classic-colors.
With that information I would probably prefer to make the default aliased, and then provide the config file for those interested in changing it to anti-aliased (at least until the patch is included). Am I correctly reading that the report has gone untouched for over a year?
from classic-colors.
@dpeterc I submitted a note to ICS with a link to your issue on bugzilla for this. I am hoping that more than one of us can get something to happen.
from classic-colors.
@dpeterc Here is what I got.
My reading of this is "we don't like the way X11 was implemented on mac", which may be true, but I don't think is a good response. Frequent round trips to the Xserver are bad, even if they are worse on mac. Am I wrong? I am happy to reply to this.
Is there actually a bug we could go after in XQuartz?
One approach to getting the fix available for users is to apply the patch in the motif hombrew formula. Is this of interest to you?
from classic-colors.
@dpeterc I will relay the message. :)
from classic-colors.
@dpeterc bugzilla on motif zone is dead. Can you post the patch here?
from classic-colors.
Download Motif from Sourceforge, edit XmRenderT.c and replace functions _XmXftDrawString2 and _XmXftDrawString.
After the fix, all Motif user interface refreshes (which use XFT fonts) on Mac OS X will be faster.
Here are the functions to replace:
void
_XmXftDrawString2(Display *display, Window window, GC gc, XftFont *font, int bpc,
#if NeedWidePrototypes
int x, int y,
#else
Position x, Position y,
#endif
char *s, int len)
{
XftDraw *draw = _XmXftDrawCreate(display, window);
XGCValues gc_val;
XftColor xftcol;
XGetGCValues(display, gc, GCForeground, &gc_val);
#ifdef FIX_1536
xftcol = _XmXftGetXftColor(display, gc_val.foreground);
#else
XColor xcol;
xcol.pixel = gc_val.foreground;
XQueryColor(display, DefaultColormap(display,
DefaultScreen(display)), &xcol);
xftcol.color.red = xcol.red;
xftcol.color.blue = xcol.blue;
xftcol.color.green = xcol.green;
xftcol.color.alpha = 0xFFFF;
#endif
switch (bpc)
{
case 1:
XftDrawStringUtf8(draw, &xftcol, font,
x, y, (XftChar8 *)s, len);
break;
case 2:
XftDrawString16(draw, &xftcol, font,
x, y, (XftChar16 *)s, len);
break;
case 4:
XftDrawString32(draw, &xftcol, font,
x, y, (XftChar32 *)s, len);
break;
default:
XmeWarning(NULL, "_XmXftDrawString(unsupported bpc)\n");
}
}
void
_XmXftDrawString(Display *display, Window window, XmRendition rend, int bpc,
#if NeedWidePrototypes
int x, int y,
#else
Position x, Position y,
#endif
char *s, int len,
#if NeedWidePrototypes
int image
#else
Boolean image
#endif
)
{
XftDraw *draw = _XmXftDrawCreate(display, window);
XftColor fg_color = _XmRendXftFG(rend);
if (image)
{
XftColor bg_color = _XmRendXftBG(rend);
XGlyphInfo ext;
ext.xOff = 0;
switch (bpc)
{
case 1:
XftTextExtentsUtf8(display, _XmRendXftFont(rend),
(FcChar8*)s, len, &ext);
break;
case 2:
XftTextExtents16(display, _XmRendXftFont(rend),
(FcChar16*)s, len, &ext);
break;
case 4:
XftTextExtents32(display, _XmRendXftFont(rend),
(FcChar32*)s, len, &ext);
break;
}
if (_XmRendBG(rend) == XmUNSPECIFIED_PIXEL)
{
XGCValues gc_val;
XGetGCValues(display, _XmRendGC(rend), GCBackground, &gc_val);
#ifdef FIX_1536
fg_color = _XmXftGetXftColor(display, gc_val.foreground);
#else
XColor xcol;
xcol.pixel = gc_val.background;
XQueryColor(display, DefaultColormapOfScreen(
DefaultScreenOfDisplay(display)), &xcol);
bg_color.pixel = xcol.pixel;
bg_color.color.red = xcol.red;
bg_color.color.green = xcol.green;
bg_color.color.blue = xcol.blue;
bg_color.color.alpha = 0xFFFF;
#endif
}
#ifdef FIX_1451
XftDrawRect(draw, &bg_color, x, y - _XmRendXftFont(rend)->ascent,
ext.xOff,
_XmRendXftFont(rend)->ascent +
_XmRendXftFont(rend)->descent);
#else
XftDrawRect(draw, &bg_color, x - 10, y - _XmRendXftFont(rend)->ascent - 10,
ext.xOff +20,
_XmRendXftFont(rend)->ascent +
_XmRendXftFont(rend)->descent + 20);
#endif
}
if (_XmRendFG(rend) == XmUNSPECIFIED_PIXEL)
{
XGCValues gc_val;
XGetGCValues(display, _XmRendGC(rend), GCForeground, &gc_val);
#ifdef FIX_1536
fg_color = _XmXftGetXftColor(display, gc_val.foreground);
#else
XColor xcol;
xcol.pixel = gc_val.foreground;
XQueryColor(display, DefaultColormapOfScreen(
DefaultScreenOfDisplay(display)), &xcol);
fg_color.pixel = xcol.pixel;
fg_color.color.red = xcol.red;
fg_color.color.green = xcol.green;
fg_color.color.blue = xcol.blue;
fg_color.color.alpha = 0xFFFF;
#endif
}
switch (bpc)
{
case 1:
XftDrawStringUtf8(draw, &fg_color, _XmRendXftFont(rend),
x, y, (XftChar8 *)s, len);
break;
case 2:
XftDrawString16(draw, &fg_color, _XmRendXftFont(rend),
x, y, (XftChar16 *)s, len);
break;
case 4:
XftDrawString32(draw, &fg_color, _XmRendXftFont(rend),
x, y, (XftChar32 *)s, len);
break;
default:
XmeWarning(NULL, "_XmXftDrawString(unsupported bpc)\n");
}
}
I also attach the complete replacement file, which contains some other small improvements and formatting fixes.
XmRenderT.c.gz
from classic-colors.
@dpeterc thanks. I'm trying to advocate for this again, either in motif or a patch for homebrew.
from classic-colors.
@justinmeiners I hope you will be successful.
It would be even better if ICS would restore Motif bugzilla or transfer its contents to sourceforge, since it contained a lot of valuable historical information, which could help in further development (or custom bug fixes).
from classic-colors.
Related Issues (20)
- openbsd gcc -O3
- Improve motif detection in configure script HOT 1
- inability to compile on macbook m1 HOT 14
- Font window doesn't work well with a tiling window manager. HOT 2
- Bad visual check in ui_draw.c HOT 3
- Provide better instructions for increasing shared memory size on Mac, in order to support MIT-SHM HOT 1
- Tool option list is not always the correct size.
- Sugestion: use extended Motif file selection box instead of the standard one HOT 6
- Loading .bmp crashes classic-colors HOT 1
- Problems compiling with pkgsrc HOT 4
- Wont build on NetBSD HOT 4
- Handle escape key to cancel tool operations
- Zoom center of focus is sometimes wrong and annoying.
- Setting "*backgroundBitmap:" in xDefaults removes all visible color options HOT 2
- Flood fill should use stack instead of a queue HOT 1
- HTML help installation should use the specified prefix in the configure/make build step.
- Assertion polygon critical values
- cc_undo_queue_trim assertion failure
- Filling a non-canvas area crashes classic-colors
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from classic-colors.