This repository has been archived on 2023-10-09. You can view files and clone it, but cannot push or open issues or pull requests.
Files
blender-archive/source/blender/windowmanager/intern/wm_subwindow.c

533 lines
12 KiB
C
Raw Normal View History

/**
* $Id: mywindow.c 9584 2007-01-03 13:45:03Z ton $
*
* ***** BEGIN GPL LICENSE BLOCK *****
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
* The Original Code is Copyright (C) 2001-2002 by NaN Holding BV.
* All rights reserved.
*
* Contributor(s): 2007 Blender Foundation (refactor)
*
* ***** END GPL LICENSE BLOCK *****
*
*
* Subwindow opengl handling.
* BTW: subwindows open/close in X11 are way too slow, tried it, and choose for my own system... (ton)
*
*/
#include <string.h>
#include "MEM_guardedalloc.h"
#include "DNA_windowmanager_types.h"
#include "DNA_screen_types.h"
#include "BLI_blenlib.h"
#include "BLI_arithb.h"
#include "BKE_global.h"
#include "BIF_gl.h"
#include "BIF_glutil.h"
#include "WM_api.h"
#include "wm_window.h"
/* wmSubWindow stored in wmWindow... but not exposed outside this C file */
/* it seems a bit redundant (area regions can store it too, but we keep it
because we can store all kind of future opengl fanciness here */
/* we use indices and array because:
- index has safety, no pointers from this C file hanging around
- fast lookups of indices with array, list would give overhead
- old code used it this way...
- keep option open to have 2 screens using same window
*/
typedef struct wmSubWindow {
struct wmSubWindow *next, *prev;
rcti winrct;
int swinid;
float viewmat[4][4], winmat[4][4];
} wmSubWindow;
/* ******************* open, free, set, get data ******************** */
/* not subwindow itself */
static void wm_subwindow_free(wmSubWindow *swin)
{
/* future fancy stuff */
}
void wm_subwindows_free(wmWindow *win)
{
wmSubWindow *swin;
for(swin= win->subwindows.first; swin; swin= swin->next)
wm_subwindow_free(swin);
BLI_freelistN(&win->subwindows);
}
void wm_subwindow_getsize(wmWindow *win, int *x, int *y)
{
if(win->curswin) {
wmSubWindow *swin= win->curswin;
*x= swin->winrct.xmax - swin->winrct.xmin + 1;
*y= swin->winrct.ymax - swin->winrct.ymin + 1;
}
}
void wm_subwindow_getorigin(wmWindow *win, int *x, int *y)
{
if(win->curswin) {
wmSubWindow *swin= win->curswin;
*x= swin->winrct.xmin;
*y= swin->winrct.ymin;
}
}
int wm_subwindow_get(wmWindow *win)
{
if(win->curswin)
return win->curswin->swinid;
return 0;
}
static wmSubWindow *swin_from_swinid(wmWindow *win, int swinid)
{
wmSubWindow *swin;
for(swin= win->subwindows.first; swin; swin= swin->next)
if(swin->swinid==swinid)
break;
return swin;
}
void wm_subwindow_set(wmWindow *win, int swinid)
{
wmSubWindow *swin= swin_from_swinid(win, swinid);
int width, height;
if(swin==NULL) {
printf("wm_subwindow_set %d: doesn't exist\n", swinid);
return;
}
win->curswin= swin;
wm_subwindow_getsize(win, &width, &height);
glViewport(swin->winrct.xmin, swin->winrct.ymin, width, height);
glScissor(swin->winrct.xmin, swin->winrct.ymin, width, height);
glMatrixMode(GL_PROJECTION);
glLoadMatrixf(&swin->winmat[0][0]);
glMatrixMode(GL_MODELVIEW);
glLoadMatrixf(&swin->viewmat[0][0]);
glFlush();
}
/* always sets pixel-precise 2D window/view matrices */
/* coords is in whole pixels. xmin = 15, xmax= 16: means window is 2 pix big */
int wm_subwindow_open(wmWindow *win, rcti *winrct)
{
wmSubWindow *swin;
int width, height;
int freewinid= 1;
for(swin= win->subwindows.first; swin; swin= swin->next)
if(freewinid <= swin->swinid)
freewinid= swin->swinid+1;
win->curswin= swin= MEM_callocN(sizeof(wmSubWindow), "swinopen");
BLI_addtail(&win->subwindows, swin);
2.5 Branch ========== * Changed wmOperatorType, removing init/exit callbacks and adding cancel callback, removed default storage in favor of properties. Defined return values for exec/invoke/modal/cancel. * Don't allocate operator on the stack, and removed operator copy for handlers. Now it frees based on return values from callbacks, and just keeps a wmOperator on the heap. Also it now registers after the operator is fully finished, to get the correct final properties. * Changed OP_get_* functions to return 1 if the property is found and 0 otherwise, gives more readable code in my opinion. Added OP_verify_* functions to quickly check if the property is available and set if it's not, that's common for exec/invoke. * Removed WM_operatortypelist_append in favor of WM_operatortype_append which takes a function pointer instead of a list, avoids macro's and duplicating code. * Fix a crash where the handler would still be used while it was freed by the operator. * Spacetypes now have operatortypes() and keymap() callbacks to abstract them a bit more. * Renamed C->curarea to C->area for consistency. Removed View3D/View2D/ SpaceIpo from bContext, seems bad to keep these. * Set context variables like window/screen/area/region to NULL again when leaving that context, instead of leaving the pointers there. * Added if(G.f & G_DEBUG) for many of the prints, makes output a bit cleaner and easier to debug. * Fixed priority of the editors/interface module in scons, would otherwise give link errors. * Added start of generic view2d api. * Added space_time with some basic drawing and a single operator to change the frame.
2008-06-11 10:10:31 +00:00
if(G.f & G_DEBUG) printf("swin %d added\n", freewinid);
swin->swinid= freewinid;
swin->winrct= *winrct;
Mat4One(swin->viewmat);
Mat4One(swin->winmat);
/* and we appy it all right away */
wm_subwindow_set(win, swin->swinid);
/* extra service */
wm_subwindow_getsize(win, &width, &height);
wmOrtho2(win, -0.375, (float)width-0.375, -0.375, (float)height-0.375);
wmLoadIdentity(win);
return swin->swinid;
}
void wm_subwindow_close(wmWindow *win, int swinid)
{
wmSubWindow *swin= swin_from_swinid(win, swinid);
if (swin) {
if (swin==win->curswin)
win->curswin= NULL;
wm_subwindow_free(swin);
BLI_remlink(&win->subwindows, swin);
MEM_freeN(swin);
}
else {
printf("wm_subwindow_close: Internal error, bad winid: %d\n", swinid);
}
}
/* pixels go from 0-99 for a 100 pixel window */
void wm_subwindow_position(wmWindow *win, int swinid, rcti *winrct)
{
wmSubWindow *swin= swin_from_swinid(win, swinid);
if(swin) {
int width, height;
swin->winrct= *winrct;
/* CRITICAL, this clamping ensures that
* the viewport never goes outside the screen
* edges (assuming the x, y coords aren't
* outside). This caused a hardware lock
* on Matrox cards if it happens.
*
* Really Blender should never _ever_ try
* to do such a thing, but just to be safe
* clamp it anyway (or fix the bScreen
* scaling routine, and be damn sure you
* fixed it). - zr (2001!)
*/
if (swin->winrct.xmax >= win->sizex)
swin->winrct.xmax= win->sizex-1;
if (swin->winrct.ymax >= win->sizey)
swin->winrct.ymax= win->sizey-1;
/* extra service */
wm_subwindow_set(win, swinid);
wm_subwindow_getsize(win, &width, &height);
wmOrtho2(win, -0.375, (float)width-0.375, -0.375, (float)height-0.375);
}
else {
printf("wm_subwindow_position: Internal error, bad winid: %d\n", swinid);
}
}
/* ---------------- WM versions of OpenGL calls, using glBlah() syntax ------------------------ */
/* ----------------- exported in WM_api.h ------------------------------------------------------ */
void wmLoadMatrix(wmWindow *win, float mat[][4])
{
if(win->curswin==NULL) return;
glLoadMatrixf(mat);
if (glaGetOneInteger(GL_MATRIX_MODE)==GL_MODELVIEW)
Mat4CpyMat4(win->curswin->viewmat, mat);
else
Mat4CpyMat4(win->curswin->winmat, mat);
}
void wmGetMatrix(wmWindow *win, float mat[][4])
{
if(win->curswin==NULL) return;
if (glaGetOneInteger(GL_MATRIX_MODE)==GL_MODELVIEW) {
Mat4CpyMat4(mat, win->curswin->viewmat);
} else {
Mat4CpyMat4(mat, win->curswin->winmat);
}
}
void wmMultMatrix(wmWindow *win, float mat[][4])
{
if(win->curswin==NULL) return;
glMultMatrixf((float*) mat);
if (glaGetOneInteger(GL_MATRIX_MODE)==GL_MODELVIEW)
glGetFloatv(GL_MODELVIEW_MATRIX, (float *)win->curswin->viewmat);
else
glGetFloatv(GL_MODELVIEW_MATRIX, (float *)win->curswin->winmat);
}
void wmGetSingleMatrix(wmWindow *win, float mat[][4])
{
if(win->curswin)
Mat4MulMat4(mat, win->curswin->viewmat, win->curswin->winmat);
}
void wmScale(wmWindow *win, float x, float y, float z)
{
if(win->curswin==NULL) return;
glScalef(x, y, z);
if (glaGetOneInteger(GL_MATRIX_MODE)==GL_MODELVIEW)
glGetFloatv(GL_MODELVIEW_MATRIX, (float *)win->curswin->viewmat);
else
glGetFloatv(GL_MODELVIEW_MATRIX, (float *)win->curswin->winmat);
}
void wmLoadIdentity(wmWindow *win)
{
if(win->curswin==NULL) return;
if (glaGetOneInteger(GL_MATRIX_MODE)==GL_MODELVIEW)
Mat4One(win->curswin->viewmat);
else
Mat4One(win->curswin->winmat);
glLoadIdentity();
}
void wmFrustum(wmWindow *win, float x1, float x2, float y1, float y2, float n, float f)
{
if(win->curswin) {
glMatrixMode(GL_PROJECTION);
glLoadIdentity();
glFrustum(x1, x2, y1, y2, n, f);
glGetFloatv(GL_PROJECTION_MATRIX, (float *)win->curswin->winmat);
glMatrixMode(GL_MODELVIEW);
}
}
void wmOrtho(wmWindow *win, float x1, float x2, float y1, float y2, float n, float f)
{
if(win->curswin) {
glMatrixMode(GL_PROJECTION);
glLoadIdentity();
glOrtho(x1, x2, y1, y2, n, f);
glGetFloatv(GL_PROJECTION_MATRIX, (float *)win->curswin->winmat);
glMatrixMode(GL_MODELVIEW);
}
}
void wmOrtho2(wmWindow *win, float x1, float x2, float y1, float y2)
{
/* prevent opengl from generating errors */
if(x1==x2) x2+=1.0;
if(y1==y2) y2+=1.0;
wmOrtho(win, x1, x2, y1, y2, -100, 100);
}
/* *************************** Framebuffer color depth, for selection codes ********************** */
static int wm_get_colordepth(void)
{
static int mainwin_color_depth= 0;
if(mainwin_color_depth==0) {
GLint r, g, b;
glGetIntegerv(GL_RED_BITS, &r);
glGetIntegerv(GL_GREEN_BITS, &g);
glGetIntegerv(GL_BLUE_BITS, &b);
mainwin_color_depth= r + g + b;
if(G.f & G_DEBUG) {
printf("Color depth r %d g %d b %d\n", (int)r, (int)g, (int)b);
glGetIntegerv(GL_AUX_BUFFERS, &r);
printf("Aux buffers: %d\n", (int)r);
}
}
return mainwin_color_depth;
}
#ifdef __APPLE__
/* apple seems to round colors to below and up on some configs */
static unsigned int index_to_framebuffer(int index)
{
unsigned int i= index;
switch(wm_get_colordepth()) {
case 12:
i= ((i & 0xF00)<<12) + ((i & 0xF0)<<8) + ((i & 0xF)<<4);
/* sometimes dithering subtracts! */
i |= 0x070707;
break;
case 15:
case 16:
i= ((i & 0x7C00)<<9) + ((i & 0x3E0)<<6) + ((i & 0x1F)<<3);
i |= 0x030303;
break;
case 24:
break;
default: // 18 bits...
i= ((i & 0x3F000)<<6) + ((i & 0xFC0)<<4) + ((i & 0x3F)<<2);
i |= 0x010101;
break;
}
return i;
}
#else
/* this is the old method as being in use for ages.... seems to work? colors are rounded to lower values */
static unsigned int index_to_framebuffer(int index)
{
unsigned int i= index;
switch(wm_get_colordepth()) {
case 8:
i= ((i & 48)<<18) + ((i & 12)<<12) + ((i & 3)<<6);
i |= 0x3F3F3F;
break;
case 12:
i= ((i & 0xF00)<<12) + ((i & 0xF0)<<8) + ((i & 0xF)<<4);
/* sometimes dithering subtracts! */
i |= 0x0F0F0F;
break;
case 15:
case 16:
i= ((i & 0x7C00)<<9) + ((i & 0x3E0)<<6) + ((i & 0x1F)<<3);
i |= 0x070707;
break;
case 24:
break;
default: // 18 bits...
i= ((i & 0x3F000)<<6) + ((i & 0xFC0)<<4) + ((i & 0x3F)<<2);
i |= 0x030303;
break;
}
return i;
}
#endif
void set_framebuffer_index_color(int index)
{
cpack(index_to_framebuffer(index));
}
int framebuffer_to_index(unsigned int col)
{
if (col==0) return 0;
switch(wm_get_colordepth()) {
case 8:
return ((col & 0xC00000)>>18) + ((col & 0xC000)>>12) + ((col & 0xC0)>>6);
case 12:
return ((col & 0xF00000)>>12) + ((col & 0xF000)>>8) + ((col & 0xF0)>>4);
case 15:
case 16:
return ((col & 0xF80000)>>9) + ((col & 0xF800)>>6) + ((col & 0xF8)>>3);
case 24:
return col & 0xFFFFFF;
default: // 18 bits...
return ((col & 0xFC0000)>>6) + ((col & 0xFC00)>>4) + ((col & 0xFC)>>2);
}
}
/* ********** END MY WINDOW ************** */
#ifdef WIN32
static int is_a_really_crappy_nvidia_card(void) {
static int well_is_it= -1;
/* Do you understand the implication? Do you? */
if (well_is_it==-1)
well_is_it= (strcmp((char*) glGetString(GL_VENDOR), "NVIDIA Corporation") == 0);
return well_is_it;
}
#endif
void myswapbuffers(void) /* XXX */
{
ScrArea *sa;
sa= G.curscreen->areabase.first;
while(sa) {
// if(sa->win_swap==WIN_BACK_OK) sa->win_swap= WIN_FRONT_OK;
// if(sa->head_swap==WIN_BACK_OK) sa->head_swap= WIN_FRONT_OK;
sa= sa->next;
}
/* HACK, some windows drivers feel they should honor the scissor
* test when swapping buffers, disable the test while swapping
* on WIN32. (namely Matrox and NVidia's new drivers around Oct 1 2001)
* - zr
*/
#ifdef WIN32
/* HACK, in some NVidia driver release some kind of
* fancy optimiziation (I presume) was put in which for
* some reason causes parts of the buffer not to be
* swapped. One way to defeat it is the following wierd
* code (which we only do for nvidia cards). This should
* be removed if NVidia fixes their drivers. - zr
*/
if (is_a_really_crappy_nvidia_card()) {
glDrawBuffer(GL_FRONT);
glBegin(GL_LINES);
glEnd();
glDrawBuffer(GL_BACK);
}
glDisable(GL_SCISSOR_TEST);
// window_swap_buffers(winlay_mainwindow);
glEnable(GL_SCISSOR_TEST);
#else
// window_swap_buffers(winlay_mainwindow);
#endif
}
/* *********************** PATTERNS ETC ***************** */
void setlinestyle(int nr) /* Move? XXX */
{
if(nr==0) {
glDisable(GL_LINE_STIPPLE);
}
else {
glEnable(GL_LINE_STIPPLE);
glLineStipple(nr, 0xAAAA);
}
}