-/* xscreensaver, Copyright (c) 1991-2012 Jamie Zawinski <jwz@jwz.org>
+/* xscreensaver, Copyright (c) 1991-2013 Jamie Zawinski <jwz@jwz.org>
*
* Permission to use, copy, modify, distribute, and sell this software and its
* documentation for any purpose is hereby granted without fee, provided that
#import <stdlib.h>
#import <stdint.h>
+#import <wchar.h>
#ifdef USE_IPHONE
# import <UIKit/UIKit.h>
#import "jwxyz-timers.h"
#import "yarandom.h"
-#ifdef USE_IPHONE
# define USE_BACKBUFFER /* must be in sync with XScreenSaverView.h */
-#endif
#undef Assert
#define Assert(C,S) do { if (!(C)) jwxyz_abort ("%s",(S)); } while(0)
}
# endif // USE_IPHONE
-#if 0
+# ifndef USE_BACKBUFFER
+ // Funny thing: As of OS X 10.9, if USE_BACKBUFFER is turned off,
+ // then this one's faster.
+
{
// Figure out this screen's colorspace, and use that for every CGImage.
//
dpy->colorspace = CGColorSpaceCreateWithPlatformColorSpace (profile);
Assert (dpy->colorspace, "unable to find colorspace");
}
-# else
+# else // USE_BACKBUFFER
// WTF? It's faster if we *do not* use the screen's colorspace!
//
dpy->colorspace = CGColorSpaceCreateDeviceRGB();
-# endif
+# endif // USE_BACKBUFFER
invalidate_drawable_cache (w);
}
#endif // USE_IPHONE
+void
+jwxyz_flush_context (Display *dpy)
+{
+ // This is only used when USE_BACKBUFFER is off.
+ CGContextFlush(dpy->main_window->cgc); // CGContextSynchronize is another possibility.
+}
jwxyz_sources_data *
display_sources_data (Display *dpy)
argb = (gc->gcv.foreground ? WhitePixel(0,0) : BlackPixel(0,0));
CGFloat x0 = wr.origin.x;
- CGFloat y0 = wr.origin.y + wr.size.height;
+ CGFloat y0 = wr.origin.y; // Y axis is refreshingly not flipped.
// It's uglier, but faster, to hoist the conditional out of the loop.
if (mode == CoordModePrevious) {
CGFloat x = x0, y = y0;
for (i = 0; i < count; i++, points++) {
x += points->x;
- y -= points->y;
+ y += points->y;
- if (0 <= x && x < w && 0 <= y && y < h) {
+ if (x >= 0 && x < w && y >= 0 && y < h) {
unsigned int *p = (unsigned int *)
((char *) data + (size_t) y * bpr + (size_t) x * 4);
*p = argb;
} else {
for (i = 0; i < count; i++, points++) {
CGFloat x = x0 + points->x;
- CGFloat y = y0 - points->y;
+ CGFloat y = y0 + points->y;
- if (0 <= x && x < w && 0 <= y && y < h) {
+ if (x >= 0 && x < w && y >= 0 && y < h) {
unsigned int *p = (unsigned int *)
((char *) data + (size_t) y * bpr + (size_t) x * 4);
*p = argb;
int x, int y, unsigned int width, unsigned int height,
BOOL foreground_p, BOOL fill_p);
+static Bool
+bitmap_context_p (Drawable d)
+{
+# ifdef USE_BACKBUFFER
+ return True;
+# else
+ // Because of the backbuffer, all iPhone Windows work like Pixmaps.
+ return d->type == PIXMAP;
+# endif
+}
+
+static void
+fill_rect_memset (void *dst, size_t dst_pitch, uint32_t fill_data,
+ size_t fill_width, size_t fill_height)
+{
+ Assert(sizeof(wchar_t) == 4, "somebody changed the ABI");
+ while (fill_height) {
+ // Would be nice if Apple used SSE/NEON in wmemset. Maybe someday.
+ wmemset (dst, fill_data, fill_width);
+ --fill_height;
+ dst = (char *) dst + dst_pitch;
+ }
+}
+
+static void *
+seek_xy (void *dst, size_t dst_pitch, unsigned x, unsigned y)
+{
+ return (char *)dst + dst_pitch * y + x * 4;
+}
+
+static unsigned int
+drawable_depth (Drawable d)
+{
+ return (d->type == WINDOW
+ ? visual_depth (NULL, NULL)
+ : d->pixmap.depth);
+}
+
+
int
XCopyArea (Display *dpy, Drawable src, Drawable dst, GC gc,
int src_x, int src_y,
// Clip rects to frames...
//
-// CGRect orig_src_rect = src_rect;
- CGRect orig_dst_rect = dst_rect;
# define CLIP(THIS,THAT,VAL,SIZE) do { \
float off = THIS##_rect.origin.VAL; \
CLIP (dst, src, x, width);
CLIP (dst, src, y, height);
+
+ // Not actually the original dst_rect, just the one before it's clipped to
+ // the src_frame.
+ CGRect orig_dst_rect = dst_rect;
+
CLIP (src, dst, x, width);
CLIP (src, dst, y, height);
# undef CLIP
- if (src_rect.size.width <= 0 || src_rect.size.height <= 0)
+ if (orig_dst_rect.size.width <= 0 || orig_dst_rect.size.height <= 0)
return 0;
+
+ // Sort-of-special case where no pixels can be grabbed from the source,
+ // and the whole destination is filled with the background color.
+ if (src_rect.size.width < 0 || src_rect.size.height < 0) {
+ src_rect.size.width = 0;
+ src_rect.size.height = 0;
+ }
NSObject *releaseme = 0;
CGImageRef cgi;
BOOL free_cgi_p = NO;
-#ifndef USE_BACKBUFFER
- // Because of the backbuffer, all iPhone Windows work like Pixmaps.
- if (src->type == PIXMAP)
-# endif
- {
+ /* If we're copying from a bitmap to a bitmap, and there's nothing funny
+ going on with clipping masks or depths or anything, optimize it by
+ just doing a memcpy instead of going through a CGI.
+ */
+ if (bitmap_context_p (src)) {
+
+ if (bitmap_context_p (dst) &&
+ gc->gcv.function == GXcopy &&
+ !gc->gcv.clip_mask &&
+ drawable_depth (src) == drawable_depth (dst)) {
+
+ Assert(!(int)src_frame.origin.x &&
+ !(int)src_frame.origin.y &&
+ !(int)dst_frame.origin.x &&
+ !(int)dst_frame.origin.y,
+ "unexpected non-zero origin");
+
+ char *src_data = CGBitmapContextGetData(src->cgc);
+ char *dst_data = CGBitmapContextGetData(dst->cgc);
+ size_t src_pitch = CGBitmapContextGetBytesPerRow(src->cgc);
+ size_t dst_pitch = CGBitmapContextGetBytesPerRow(dst->cgc);
+
+ // Int to float and back again. It's not very safe, but it seems to work.
+ int src_x0 = src_rect.origin.x;
+ int dst_x0 = dst_rect.origin.x;
+
+ // Flip the Y-axis a second time.
+ int src_y0 = (src_frame.origin.y + src_frame.size.height -
+ src_rect.size.height - src_rect.origin.y);
+ int dst_y0 = (dst_frame.origin.y + dst_frame.size.height -
+ dst_rect.size.height - dst_rect.origin.y);
+
+ unsigned width0 = (int) src_rect.size.width;
+ unsigned height0 = (int) src_rect.size.height;
+
+ Assert((int)src_rect.size.width == (int)dst_rect.size.width ||
+ (int)src_rect.size.height == (int)dst_rect.size.height,
+ "size mismatch");
+ {
+ char *src_data0 = seek_xy(src_data, src_pitch, src_x0, src_y0);
+ char *dst_data0 = seek_xy(dst_data, dst_pitch, dst_x0, dst_y0);
+ size_t src_pitch0 = src_pitch;
+ size_t dst_pitch0 = dst_pitch;
+ size_t bytes = width0 * 4;
+
+ if (src == dst && dst_y0 > src_y0) {
+ // Copy upwards if the areas might overlap.
+ src_data0 += src_pitch0 * (height0 - 1);
+ dst_data0 += dst_pitch0 * (height0 - 1);
+ src_pitch0 = -src_pitch0;
+ dst_pitch0 = -dst_pitch0;
+ }
+
+ size_t lines0 = height0;
+ while (lines0) {
+ // memcpy is an alias for memmove on OS X.
+ memmove(dst_data0, src_data0, bytes);
+ src_data0 += src_pitch0;
+ dst_data0 += dst_pitch0;
+ --lines0;
+ }
+ }
+
+ if (clipped) {
+ int orig_dst_x = orig_dst_rect.origin.x;
+ int orig_dst_y = (dst_frame.origin.y + dst_frame.size.height -
+ orig_dst_rect.origin.y - orig_dst_rect.size.height);
+ int orig_width = orig_dst_rect.size.width;
+ int orig_height = orig_dst_rect.size.height;
+
+ Assert (orig_dst_x >= 0 &&
+ orig_dst_x + orig_width <= (int) dst_frame.size.width &&
+ orig_dst_y >= 0 &&
+ orig_dst_y + orig_height <= (int) dst_frame.size.height,
+ "wrong dimensions");
+
+ if (orig_dst_y < dst_y0) {
+ fill_rect_memset (seek_xy (dst_data, dst_pitch,
+ orig_dst_x, orig_dst_y), dst_pitch,
+ gc->gcv.background, orig_width,
+ dst_y0 - orig_dst_y);
+ }
+
+ if (orig_dst_y + orig_height > dst_y0 + height0) {
+ fill_rect_memset (seek_xy (dst_data, dst_pitch, orig_dst_x,
+ dst_y0 + height0),
+ dst_pitch,
+ gc->gcv.background, orig_width,
+ orig_dst_y + orig_height - dst_y0 - height0);
+ }
+
+ if (orig_dst_x < dst_x0) {
+ fill_rect_memset (seek_xy (dst_data, dst_pitch, orig_dst_x, dst_y0),
+ dst_pitch, gc->gcv.background,
+ dst_x0 - orig_dst_x, height0);
+ }
+
+ if (dst_x0 + width0 < orig_dst_x + orig_width) {
+ fill_rect_memset (seek_xy (dst_data, dst_pitch, dst_x0 + width0,
+ dst_y0),
+ dst_pitch, gc->gcv.background,
+ orig_dst_x + orig_width - dst_x0 - width0,
+ height0);
+ }
+ }
+
+ invalidate_drawable_cache (dst);
+ return 0;
+ }
+
// If we are copying from a Pixmap to a Pixmap or Window, we must first
// copy the bits to an intermediary CGImage object, then copy that to the
unsigned long plane_mask, int format)
{
const unsigned char *data = 0;
- int depth, ibpp, ibpl, alpha_first_p;
+ int depth, ibpp, ibpl;
+ enum { RGBA, ARGB, BGRA } src_format; // As bytes.
# ifndef USE_BACKBUFFER
NSBitmapImageRep *bm = 0;
# endif
depth = (d->type == PIXMAP
? d->pixmap.depth
: 32);
- // If it's a pixmap, we created it with kCGImageAlphaNoneSkipFirst.
- // If it's an iPhone window, it's the other way around.
- alpha_first_p = (d->type == PIXMAP);
+ // We create pixmaps and iPhone backbuffers with kCGImageAlphaNoneSkipFirst.
+ src_format = BGRA; // #### Should this be ARGB on PPC?
ibpp = CGBitmapContextGetBitsPerPixel (cgc);
ibpl = CGBitmapContextGetBytesPerRow (cgc);
data = CGBitmapContextGetData (cgc);
// get the bits (desired sub-rectangle) out of the NSView
NSRect nsfrom;
nsfrom.origin.x = x;
- nsfrom.origin.y = y;
+// nsfrom.origin.y = y;
+ nsfrom.origin.y = d->frame.size.height - height - y;
nsfrom.size.width = width;
nsfrom.size.height = height;
bm = [[NSBitmapImageRep alloc] initWithFocusedViewRect:nsfrom];
depth = 32;
- alpha_first_p = ([bm bitmapFormat] & NSAlphaFirstBitmapFormat);
+ src_format = ([bm bitmapFormat] & NSAlphaFirstBitmapFormat) ? ARGB : RGBA;
ibpp = [bm bitsPerPixel];
ibpl = [bm bytesPerRow];
data = [bm bitmapData];
const unsigned char *iline2 = iline;
unsigned char *oline2 = oline;
- if (alpha_first_p) // ARGB
+ switch (src_format) {
+ case ARGB:
for (xx = 0; xx < width; xx++) {
unsigned char a = (ibpp == 32 ? (*iline2++) : 0xFF);
unsigned char r = *iline2++;
*((uint32_t *) oline2) = pixel;
oline2 += 4;
}
- else // RGBA
+ break;
+ case RGBA:
for (xx = 0; xx < width; xx++) {
unsigned char r = *iline2++;
unsigned char g = *iline2++;
*((uint32_t *) oline2) = pixel;
oline2 += 4;
}
+ break;
+ case BGRA:
+ for (xx = 0; xx < width; xx++) {
+ unsigned char b = *iline2++;
+ unsigned char g = *iline2++;
+ unsigned char r = *iline2++;
+ unsigned char a = (ibpp == 32 ? (*iline2++) : 0xFF);
+ uint32_t pixel = ((a << 24) |
+ (r << 16) |
+ (g << 8) |
+ (b << 0));
+ *((uint32_t *) oline2) = pixel;
+ oline2 += 4;
+ }
+ break;
+ default:
+ abort();
+ break;
+ }
oline += obpl;
iline += ibpl;
width * 4, /* bpl */
dpy->colorspace,
// Without this, it returns 0...
- kCGImageAlphaNoneSkipFirst
+ (kCGImageAlphaNoneSkipFirst |
+ kCGBitmapByteOrder32Host)
);
Assert (p->cgc, "could not create CGBitmapContext");
return p;
width * 4, /* bpl */
dpy->colorspace,
// Without this, it returns 0...
- kCGImageAlphaNoneSkipFirst
+ (kCGImageAlphaNoneSkipFirst |
+ kCGBitmapByteOrder32Host)
);
Assert (p2->cgc, "could not create CGBitmapContext");
XCharStruct *min = &f->min_bounds;
XCharStruct *max = &f->max_bounds;
-#define CEIL(F) ((F) < 0 ? floor(F) : ceil(F))
+#define CEIL(F) ((F) < 0 ? floor(F) : ceil(F))
f->fid = fid;
f->min_char_or_byte2 = first;
advancement.x = adv.width;
advancement.y = adv.height;
- // Seems to be clipping by a pixel or two. Add a margin to be safe.
- bbox.origin.x -= 2;
- bbox.origin.y -= 2;
- bbox.size.width += 4;
- bbox.size.height += 4;
+ /* A bug that existed was that the GL FPS display was truncating
+ characters slightly: commas looked like periods.
+
+ At one point, I believed the bounding box was being rounded
+ wrong and we needed to add padding to it here.
+
+ I think what was actually going on was, I was computing rbearing
+ wrong. Also there was an off-by-one error in texfont.c, displaying
+ too little of the bitmap.
+
+ Adding arbitrarily large padding to the bbox is fine in fontglide
+ and FPS display, but screws up BSOD. Increasing bbox width makes
+ inverted text print too wide; decreasing origin makes characters
+ clip.
+
+ I think that all 3 states are correct now with the new lbearing
+ computation plus the texfont fix.
+ */
+# if 0
+ double kludge = 2;
+ bbox.origin.x -= kludge;
+ bbox.origin.y -= kludge;
+ bbox.size.width += kludge;
+ bbox.size.height += kludge;
+# endif
}
# endif // USE_IPHONE
cs->ascent = CEIL (bbox.origin.y) + CEIL (bbox.size.height);
cs->descent = CEIL(-bbox.origin.y);
- cs->lbearing = CEIL (bbox.origin.x);
- cs->rbearing = CEIL (bbox.origin.x) + CEIL (bbox.size.width);
+ cs->lbearing = floor (bbox.origin.x);
+// cs->rbearing = CEIL (bbox.origin.x) + CEIL (bbox.size.width);
+ cs->rbearing = CEIL (bbox.origin.x + bbox.size.width) - cs->lbearing;
cs->width = CEIL (advancement.x);
- Assert (cs->rbearing - cs->lbearing == CEIL(bbox.size.width),
- "bbox w wrong");
+// Assert (cs->rbearing - cs->lbearing == CEIL(bbox.size.width),
+// "bbox w wrong");
Assert (cs->ascent + cs->descent == CEIL(bbox.size.height),
"bbox h wrong");
#if 0
fprintf(stderr, " %3d %c: w=%3d lb=%3d rb=%3d as=%3d ds=%3d "
- " bb=%3d x %3d @ %3d %3d adv=%3d %3d\n",
+ " bb=%5.1f x %5.1f @ %5.1f %5.1f adv=%5.1f %5.1f\n",
i, i, cs->width, cs->lbearing, cs->rbearing,
cs->ascent, cs->descent,
- (int) bbox.size.width, (int) bbox.size.height,
- (int) bbox.origin.x, (int) bbox.origin.y,
- (int) advancement.x, (int) advancement.y);
+ bbox.size.width, bbox.size.height,
+ bbox.origin.x, bbox.origin.y,
+ advancement.x, advancement.y);
#endif
}
projects / xscreensaver / blobdiff