March 2009 - Xquartz-changes - lists.macosforge.org

[331] trunk/pkg/README.2.3.3.txt
by source_changes＠macosforge.org 31 Mar '09

31 Mar '09

Revision: 331 http://trac.macosforge.org/projects/xquartz/changeset/331 Author: jeremyhu(a)freedesktop.org Date: 2009-03-31 13:54:55 -0700 (Tue, 31 Mar 2009) Log Message: ----------- Updated 2.3.3 notes Modified Paths: -------------- trunk/pkg/README.2.3.3.txt Modified: trunk/pkg/README.2.3.3.txt =================================================================== --- trunk/pkg/README.2.3.3.txt 2009-03-31 06:35:49 UTC (rev 330) +++ trunk/pkg/README.2.3.3.txt 2009-03-31 20:54:55 UTC (rev 331) @@ -34,7 +34,7 @@ * xcb-proto 1.4 * xf86rushproto 1.1.2 * lib: - * AppleSGLX 55 + * AppleSGLX 56 * Support the same version of OpenGL as OpenGL.framework (2.1) * GLSL and other features now supported * Many extensions from OpenGL.framework are now available @@ -60,9 +60,10 @@ * pixman 0.14.0 * xtrans 1.2.3 * server: - * Xquartz fixes from xorg-server-1.4.2-apple37 + * Xquartz fixes from xorg-server-1.4.2-apple38 * DRI Updates for new libGL * Fix CapsLock + * Fixed mouse tracking for games like Quake2, Halflife, etc in wine. == Changes in 2.3.2.1 == * All changes in 2.3.2 plus:

1 0

[330] AppleSGLX/trunk/Makefile
by source_changes＠macosforge.org 31 Mar '09

31 Mar '09

Revision: 330 http://trac.macosforge.org/projects/xquartz/changeset/330 Author: jeremyhu(a)freedesktop.org Date: 2009-03-30 23:35:49 -0700 (Mon, 30 Mar 2009) Log Message: ----------- Fix parallel make Modified Paths: -------------- AppleSGLX/trunk/Makefile Modified: AppleSGLX/trunk/Makefile =================================================================== --- AppleSGLX/trunk/Makefile 2009-03-31 03:43:31 UTC (rev 329) +++ AppleSGLX/trunk/Makefile 2009-03-31 06:35:49 UTC (rev 330) @@ -12,7 +12,7 @@ LN=ln RM=rm -INCLUDE=-I. -Iinclude -Iinclude/internal -DGLX_ALIAS_UNSUPPORTED -F/System/Library/Frameworks/OpenGL.framework -I$(INSTALL_DIR)/include -I$(X11_DIR)/include +INCLUDE=-I. -Iinclude -Iinclude/internal -DGLX_ALIAS_UNSUPPORTED -I$(INSTALL_DIR)/include -I$(X11_DIR)/include COMPILE=$(CC) $(INCLUDE) $(GL_CFLAGS) -c #The directory with the final binaries. @@ -52,7 +52,7 @@ apple_xgl_api.o: apple_xgl_api.h apple_xgl_api.c apple_xgl_api_stereo.c include/GL/gl.h apple_xgl_api_read.o: apple_xgl_api_read.h apple_xgl_api_read.c apple_xgl_api.h include/GL/gl.h apple_xgl_api_viewport.o: apple_xgl_api_viewport.h apple_xgl_api_viewport.c apple_xgl_api.h include/GL/gl.h -apple_xgl_api_stereo.o: apple_xgl_api_stereo.h apple_xgl_api_stereo.c +apple_xgl_api_stereo.o: apple_xgl_api_stereo.h apple_xgl_api_stereo.c apple_xgl_api.h include/GL/gl.h glcontextmodes.o: glcontextmodes.c glcontextmodes.h include/GL/gl.h glxext.o: glxext.c include/GL/gl.h glxreply.o: glxreply.c include/GL/gl.h

1 0

[329] AppleSGLX/tags/AppleSGLX-56/
by source_changes＠macosforge.org 31 Mar '09

31 Mar '09

Revision: 329 http://trac.macosforge.org/projects/xquartz/changeset/329 Author: jeremyhu(a)freedesktop.org Date: 2009-03-30 20:43:31 -0700 (Mon, 30 Mar 2009) Log Message: ----------- Tagging AppleSGLX-56 Added Paths: ----------- AppleSGLX/tags/AppleSGLX-56/

1 0

[328] AppleSGLX/trunk/OVERALL_DESIGN
by source_changes＠macosforge.org 27 Mar '09

27 Mar '09

Revision: 328 http://trac.macosforge.org/projects/xquartz/changeset/328 Author: gstaplin(a)apple.com Date: 2009-03-27 13:48:37 -0700 (Fri, 27 Mar 2009) Log Message: ----------- Update the OVERALL_DESIGN to reflect the current way the tests are built and where they are stored. Modified Paths: -------------- AppleSGLX/trunk/OVERALL_DESIGN Modified: AppleSGLX/trunk/OVERALL_DESIGN =================================================================== --- AppleSGLX/trunk/OVERALL_DESIGN 2009-03-27 20:48:13 UTC (rev 327) +++ AppleSGLX/trunk/OVERALL_DESIGN 2009-03-27 20:48:37 UTC (rev 328) @@ -1,4 +1,3 @@ - A lot of the code is automatically generated. The following are generated based on specs/gl.spec and specs/enum.spec (from OpenGL.org) @@ -6,15 +5,14 @@ apple_xgl_api.h apple_xgl_api.c exports.list -include/GL/gl.h (this probably might as well just be an include of the OpenGL framework GL) +include/GL/gl.h include/GL/glext.h (includes the OpenGL framework glext.h) The gen_code.tcl script is what executes the various gen_*.tcl scripts to produce those. You will need Tcl 8.5 for the gen_code.tcl script. -The tests/ directory contains some tests that are built in builds/. -(TODO move the libGL.dylib and libGL.1.2.dylib, libGL.1.3.dylib creation into builds) +The tests/ directory contains some tests that are built in testbuilds. -The tests built in builds don't require installation of the library. +The tests built in testbuilds don't require installation of the library.

1 0

[327] AppleSGLX/trunk/RELEASE_NOTES
by source_changes＠macosforge.org 27 Mar '09

27 Mar '09

Revision: 327 http://trac.macosforge.org/projects/xquartz/changeset/327 Author: gstaplin(a)apple.com Date: 2009-03-27 13:48:13 -0700 (Fri, 27 Mar 2009) Log Message: ----------- Update the RELEASE_NOTES to reflect the current reality. Modified Paths: -------------- AppleSGLX/trunk/RELEASE_NOTES Modified: AppleSGLX/trunk/RELEASE_NOTES =================================================================== --- AppleSGLX/trunk/RELEASE_NOTES 2009-03-27 19:02:55 UTC (rev 326) +++ AppleSGLX/trunk/RELEASE_NOTES 2009-03-27 20:48:13 UTC (rev 327) @@ -52,8 +52,11 @@ contexts requires a similar visual or GLXFBConfig be used in the creation of a shared context. It's best if you specify the same visual. This is due to a CGL design difference, and not something -that is easily worked around. +that is easily worked around. UPDATE: some changes made seem to +help resolve this issue in many cases, so you may be able to use a +shared context without this restriction. + o Indirect The X server supports indirect fairly well, so OpenGL applications

1 0

[326] AppleSGLX/trunk
by source_changes＠macosforge.org 27 Mar '09

27 Mar '09

Revision: 326 http://trac.macosforge.org/projects/xquartz/changeset/326 Author: gstaplin(a)apple.com Date: 2009-03-27 12:02:55 -0700 (Fri, 27 Mar 2009) Log Message: ----------- Fix the surface destruction issue more cleanly with a slightly better design. The test case still passes, and this should prevent a memory leak pattern that the previous fix would have suffered from. Tested with a few apps, and triangle_glx_destroy_relation (the specific test case for this). Modified Paths: -------------- AppleSGLX/trunk/apple_glx_context.c AppleSGLX/trunk/apple_glx_surface.c Modified: AppleSGLX/trunk/apple_glx_context.c =================================================================== --- AppleSGLX/trunk/apple_glx_context.c 2009-03-26 07:53:24 UTC (rev 325) +++ AppleSGLX/trunk/apple_glx_context.c 2009-03-27 19:02:55 UTC (rev 326) @@ -572,14 +572,7 @@ * If there are references to it, then it's probably made * current in another context. */ - d->destroy(d); - - /* - * FIXME: We retain 2 references to surfaces to prevent - * premature destruction. See also: the comment in the - * make current path. - */ - d->destroy(d); + d->destroy(d); } } } Modified: AppleSGLX/trunk/apple_glx_surface.c =================================================================== --- AppleSGLX/trunk/apple_glx_surface.c 2009-03-26 07:53:24 UTC (rev 325) +++ AppleSGLX/trunk/apple_glx_surface.c 2009-03-27 19:02:55 UTC (rev 326) @@ -101,16 +101,23 @@ fprintf(stderr, "xp_destroy_surface error: %d\n", (int)error); } - /* - * Warning: this causes other routines to be called (potentially) - * from surface_notify_handler. It's probably best to not have - * any locks at this point locked. + /* + * Check if this surface destroy came from the surface being destroyed + * on the server. If s->pending_destroy is true, then it did, and + * we don't want to try to destroy the surface on the server. */ - XAppleDRIDestroySurface(d->display, DefaultScreen(d->display), - d->drawable); - - apple_glx_diagnostic("%s: destroyed a surface for drawable 0x%lx uid %u\n", - __func__, d->drawable, s->uid); + if(!s->pending_destroy) { + /* + * Warning: this causes other routines to be called (potentially) + * from surface_notify_handler. It's probably best to not have + * any locks at this point locked. + */ + XAppleDRIDestroySurface(d->display, DefaultScreen(d->display), + d->drawable); + + apple_glx_diagnostic("%s: destroyed a surface for drawable 0x%lx uid %u\n", + __func__, d->drawable, s->uid); + } } /* Return true if an error occured. */ @@ -193,6 +200,18 @@ if(d) { d->types.surface.pending_destroy = true; d->release(d); + /* + * We release 2 references to the surface. One was acquired by + * the find, and the other was leftover from a context, or + * the surface being displayed, so the destroy() will decrease it + * once more. + * + * If the surface is in a context, it will take one d->destroy(d); + * to actually destroy it when the pending_destroy is processed + * by a glViewport callback (see apple_glx_context_update()). + */ + d->destroy(d); + d->unlock(d); } }

1 0

[325] AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c
by source_changes＠macosforge.org 26 Mar '09

26 Mar '09

Revision: 325 http://trac.macosforge.org/projects/xquartz/changeset/325 Author: gstaplin(a)apple.com Date: 2009-03-26 00:53:24 -0700 (Thu, 26 Mar 2009) Log Message: ----------- Improve the new triangle_glx_destroy_relation test, so that we can test some more state interactively. This is useful with LIBGL_DIAGNOSTIC. Modified Paths: -------------- AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c Modified: AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c =================================================================== --- AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c 2009-03-26 07:51:57 UTC (rev 324) +++ AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c 2009-03-26 07:53:24 UTC (rev 325) @@ -25,23 +25,28 @@ void event_loop(Display *dpy) { XEvent event; - + XWindowAttributes wattr; + while(1) { XNextEvent(dpy, &event); switch(event.type) { + case ButtonPress: + printf("event.xbutton.window 0x%lx\n", + event.xbutton.window); + XDestroyWindow(dpy, event.xbutton.window); + break; + case Expose: glXMakeCurrent(dpy, event.xexpose.window, ctx); + XGetWindowAttributes(dpy, event.xexpose.window, &wattr); + resize(dpy, event.xexpose.window, wattr.width, wattr.height); draw(dpy, event.xexpose.window); break; - - case KeyPress: { - XWindowAttributes wattr; - + + case KeyPress: { glXMakeCurrent(dpy, win2, ctx); - XGetWindowAttributes(dpy, win2, &wattr); - resize(dpy, win2, wattr.width, wattr.height); draw(dpy, win2); XFlush(dpy); @@ -97,7 +102,8 @@ attr.background_pixel = 0; attr.border_pixel = 0; attr.colormap = XCreateColormap(dpy, root, visinfo->visual, AllocNone); - attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask; + attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask + | ButtonPressMask; win = XCreateWindow(dpy, root, /*x*/ 0, /*y*/ 0, /*width*/ 400, /*height*/ 400, @@ -113,6 +119,8 @@ CWBackPixel | CWBorderPixel | CWColormap | CWEventMask, &attr); + printf("win 0x%lx win2 0x%lx\n", win, win2); + ctx = glXCreateContext(dpy, visinfo, NULL, True); if(!ctx) { fprintf(stderr, "error: glXCreateContext failed!\n");

1 0

[324] AppleSGLX/trunk/apple_glx_context.c
by source_changes＠macosforge.org 26 Mar '09

26 Mar '09

Revision: 324 http://trac.macosforge.org/projects/xquartz/changeset/324 Author: gstaplin(a)apple.com Date: 2009-03-26 00:51:57 -0700 (Thu, 26 Mar 2009) Log Message: ----------- Work around an issue with black windows that occurred with some apps when using the new libGL. This behavior should be compatible with the old libGL, though it's not ideal. See the comments I added with this commit about a unified rendering layer for more info. Modified Paths: -------------- AppleSGLX/trunk/apple_glx_context.c Modified: AppleSGLX/trunk/apple_glx_context.c =================================================================== --- AppleSGLX/trunk/apple_glx_context.c 2009-03-26 07:18:13 UTC (rev 323) +++ AppleSGLX/trunk/apple_glx_context.c 2009-03-26 07:51:57 UTC (rev 324) @@ -360,6 +360,27 @@ /* The drawable is referenced once by apple_glx_surface_create. */ + /* + * FIXME: We actually need 2 references to prevent premature surface + * destruction. The problem is that the surface gets destroyed in + * the case of the context being reused for another window, and + * we then lose the surface contents. Wait for destruction of a + * window to destroy a surface. + * + * Note: this may leave around surfaces we don't want around, if + * say we are using X for raster drawing after OpenGL rendering, + * but it will be compatible with the old libGL's behavior. + * + * Someday the X11 and OpenGL rendering must be unified at some + * layer. I suspect we can do that via shared memory and + * multiple threads in the X server (1 for each context created + * by a client). This would also allow users to render from + * multiple clients to the same OpenGL surface. In fact it could + * all be OpenGL. + * + */ + newagd->reference(newagd); + /* Save the new drawable with the context structure. */ ac->drawable = newagd; } else { @@ -552,7 +573,14 @@ * current in another context. */ d->destroy(d); - } + + /* + * FIXME: We retain 2 references to surfaces to prevent + * premature destruction. See also: the comment in the + * make current path. + */ + d->destroy(d); + } } }

1 0

[323] AppleSGLX/trunk/tests
by source_changes＠macosforge.org 26 Mar '09

26 Mar '09

Revision: 323 http://trac.macosforge.org/projects/xquartz/changeset/323 Author: gstaplin(a)apple.com Date: 2009-03-26 00:18:13 -0700 (Thu, 26 Mar 2009) Log Message: ----------- Add a test that demonstrates the black surface problem some users are experiencing. Modified Paths: -------------- AppleSGLX/trunk/tests/tests.mk AppleSGLX/trunk/tests/triangle_glx/triangle_glx.mk Added Paths: ----------- AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c Modified: AppleSGLX/trunk/tests/tests.mk =================================================================== --- AppleSGLX/trunk/tests/tests.mk 2009-03-21 00:36:03 UTC (rev 322) +++ AppleSGLX/trunk/tests/tests.mk 2009-03-26 07:18:13 UTC (rev 323) @@ -34,5 +34,6 @@ $(TEST_BUILD_DIR)/glthreads \ $(TEST_BUILD_DIR)/triangle_glx_surface \ $(TEST_BUILD_DIR)/triangle_glx_surface-2 \ - $(TEST_BUILD_DIR)/triangle_glx_withdraw_remap + $(TEST_BUILD_DIR)/triangle_glx_withdraw_remap \ + $(TEST_BUILD_DIR)/triangle_glx_destroy_relation Modified: AppleSGLX/trunk/tests/triangle_glx/triangle_glx.mk =================================================================== --- AppleSGLX/trunk/tests/triangle_glx/triangle_glx.mk 2009-03-21 00:36:03 UTC (rev 322) +++ AppleSGLX/trunk/tests/triangle_glx/triangle_glx.mk 2009-03-26 07:18:13 UTC (rev 323) @@ -9,3 +9,6 @@ $(TEST_BUILD_DIR)/triangle_glx_withdraw_remap: tests/triangle_glx/triangle_glx_withdraw_remap.c $(LIBGL) $(CC) tests/triangle_glx/triangle_glx_withdraw_remap.c -Iinclude -o $(TEST_BUILD_DIR)/triangle_glx_withdraw_remap $(LINK_TEST) + +$(TEST_BUILD_DIR)/triangle_glx_destroy_relation: tests/triangle_glx/triangle_glx_destroy_relation.c $(LIBGL) + $(CC) tests/triangle_glx/triangle_glx_destroy_relation.c -Iinclude -o $(TEST_BUILD_DIR)/triangle_glx_destroy_relation $(LINK_TEST) Added: AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c =================================================================== --- AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c (rev 0) +++ AppleSGLX/trunk/tests/triangle_glx/triangle_glx_destroy_relation.c 2009-03-26 07:18:13 UTC (rev 323) @@ -0,0 +1,130 @@ +#include <stdio.h> +#include <stdlib.h> +#include <X11/Xlib.h> +#include <GL/glx.h> + +Window root, win, win2; +GLXContext ctx; + +void draw(Display *dpy, Window w) { + glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT); + glLoadIdentity(); + glColor3f(0.5f, 0.5f, 1.0f); + glBegin(GL_TRIANGLES); + glVertex3f( 0.0f, 1.0f, 0.0f); + glVertex3f(-1.0f,-1.0f, 0.0f); + glVertex3f( 1.0f,-1.0f, 0.0f); + glEnd(); + glXSwapBuffers(dpy, w); +} + +void resize(Display *dpy, Window w, int width, int height) { + glViewport(0, 0, width, height); + draw(dpy, w); +} + +void event_loop(Display *dpy) { + XEvent event; + + while(1) { + XNextEvent(dpy, &event); + + switch(event.type) { + case Expose: + glXMakeCurrent(dpy, event.xexpose.window, ctx); + draw(dpy, event.xexpose.window); + break; + + case KeyPress: { + XWindowAttributes wattr; + + glXMakeCurrent(dpy, win2, ctx); + + XGetWindowAttributes(dpy, win2, &wattr); + + resize(dpy, win2, wattr.width, wattr.height); + draw(dpy, win2); + XFlush(dpy); + } + break; + + case ConfigureNotify: + glXMakeCurrent(dpy, event.xconfigure.window, ctx); + resize(dpy, event.xconfigure.window, event.xconfigure.width, + event.xconfigure.height); + break; + } + } +} + +int main() { + Display *dpy; + int attrib[] = { GLX_RGBA, + GLX_RED_SIZE, 8, + GLX_GREEN_SIZE, 8, + GLX_BLUE_SIZE, 8, + GLX_DEPTH_SIZE, 24, + GLX_DOUBLEBUFFER, + None }; + int eventbase, errorbase; + int screen; + XVisualInfo *visinfo; + XSetWindowAttributes attr; + + + dpy = XOpenDisplay(NULL); + + if(NULL == dpy) { + fprintf(stderr, "error: unable to open display!\n"); + return EXIT_FAILURE; + } + + if(!glXQueryExtension(dpy, &eventbase, &errorbase)) { + fprintf(stderr, "GLX is not available!\n"); + return EXIT_FAILURE; + } + + screen = DefaultScreen(dpy); + root = RootWindow(dpy, screen); + + visinfo = glXChooseVisual(dpy, screen, attrib); + + if(!visinfo) { + fprintf(stderr, "error: couldn't get an RGBA, double-buffered visual!\n"); + return EXIT_FAILURE; + } + + attr.background_pixel = 0; + attr.border_pixel = 0; + attr.colormap = XCreateColormap(dpy, root, visinfo->visual, AllocNone); + attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask; + + win = XCreateWindow(dpy, root, /*x*/ 0, /*y*/ 0, + /*width*/ 400, /*height*/ 400, + 0, visinfo->depth, InputOutput, + visinfo->visual, + CWBackPixel | CWBorderPixel | CWColormap | CWEventMask, + &attr); + + win2 = XCreateWindow(dpy, win, 0, 0, + 100, 100, + 0, visinfo->depth, InputOutput, + visinfo->visual, + CWBackPixel | CWBorderPixel | CWColormap | CWEventMask, + &attr); + + ctx = glXCreateContext(dpy, visinfo, NULL, True); + if(!ctx) { + fprintf(stderr, "error: glXCreateContext failed!\n"); + return EXIT_FAILURE; + } + + XMapWindow(dpy, win); + XMapWindow(dpy, win2); + + glXMakeCurrent(dpy, win, ctx); + + event_loop(dpy); + + return EXIT_SUCCESS; +}

1 0

[322] AppleSGLX/trunk/apple_glx.c
by source_changes＠macosforge.org 21 Mar '09

21 Mar '09

Revision: 322 http://trac.macosforge.org/projects/xquartz/changeset/322 Author: gstaplin(a)apple.com Date: 2009-03-20 17:36:03 -0700 (Fri, 20 Mar 2009) Log Message: ----------- Reorder some code so that each display connection gets initialized. Modified Paths: -------------- AppleSGLX/trunk/apple_glx.c Modified: AppleSGLX/trunk/apple_glx.c =================================================================== --- AppleSGLX/trunk/apple_glx.c 2009-03-20 22:37:15 UTC (rev 321) +++ AppleSGLX/trunk/apple_glx.c 2009-03-21 00:36:03 UTC (rev 322) @@ -112,6 +112,12 @@ int eventBase, errorBase; int major, minor, patch; + if(!XAppleDRIQueryExtension(dpy, &eventBase, &errorBase)) + return true; + + if(!XAppleDRIQueryVersion(dpy, &major, &minor, &patch)) + return true; + if(initialized) return false; @@ -125,14 +131,9 @@ libgl_handle = dlopen(OPENGL_LIB_PATH, RTLD_LAZY); (void)apple_glx_get_client_id(); - if(!XAppleDRIQueryExtension(dpy, &eventBase, &errorBase)) - return true; - - if(!XAppleDRIQueryVersion(dpy, &major, &minor, &patch)) - return true; - XAppleDRISetSurfaceNotifyHandler(surface_notify_handler); + /* This should really be per display. */ dri_event_base = eventBase; initialized = true;

1 0