summaryrefslogtreecommitdiffstats
path: root/src
Commit message (Collapse)AuthorAgeFilesLines
* Android/NewtVersionActivity: Dump avail. GLCaps / Use stderr, since log ↵Sven Gothel2013-01-192-8/+28
| | | | would cap 'em (too long).
* TestSwingAWTRobotUsageBeforeJOGLInitBug411: Fix 'chicken egg' init problem ↵Sven Gothel2013-01-191-1/+12
| | | | | | w/ AWTRobot, i.e. NativeWindow not yet initialized. Regression since: 50f997557b91a2f014ef0c2ea848c5c326d0cfb2
* Adapt to change 5060b728b26a89a14367f8fca9f4eb7968f719ca: ↵Sven Gothel2013-01-191-3/+5
| | | | ShaderState.uniform(..) returns false if no location available.
* NEWT/Android: Full Lifecycle for WindowDriver; Using static ViewGroup; ↵Sven Gothel2013-01-195-178/+252
| | | | | | | | | | | | | | | | AWTRobotUtil: More tolerant for non AWT env.; Fix adb-launch-* - NEWT/Android WindowDriver - Full Lifecycle, remove refs on closeNative() - Respect isFullscreen() - Using static ViewGroup if available and surface not ready, allows running from main() - AWTRobotUtil: More tolerant for non AWT env. - Check for NEWT first - Only use AWT iff available, which allows running on Android - Fix adb-launch-* - Launch main/junit tests
* Fix Bug 669: Recursive GLContext makeCurrent()/release()Sven Gothel2013-01-185-38/+364
| | | | | | | | | | | | | | | | | | | | | | Culprit: GLContext's makeCurrent() didn't clear the boolean flag 'unlockContextAndSurface' in case the context is already current (-> recursion). Above case was detected within a code block tailed by a finally block, which acted on mentioned flag, i.e. called lock.unlock() and hence decremented the lock count even though the method return w/ successful state. Fixed. Added debug code: GLContext.release() debug code (DEBUG | TRACE_SWITCH), recording stack trace of last release() call, which is dumped in case no current was current. Added 2 unit tests: - Simple recursive GLContext makeCurrent()/release() from within GLEventListener's display(). Test also validates lock count and lock ownership. - GLAutoDrawable display() of another GLAutoDrawable from within GLEventListener's display(..).
* Sync LauncherUtil w/ GlueGen commit 1b9f0739ecc25105384b557afa698c42e08d4cc6 ↵Sven Gothel2013-01-181-13/+57
| | | | (adding main-cmdline-args)
* Merge remote-tracking branch 'xranby/NEWT-input'Sven Gothel2013-01-161-1/+2
|\
| * LinuxMouseTracker: Fix inverted Y-axis & missing button release.Xerxes Rånby2013-01-151-1/+2
| | | | | | | | Signed-off-by: Xerxes Rånby <[email protected]>
* | ShaderState: Reduce hash-map utilization at shader switch (reset ↵Sven Gothel2013-01-151-44/+59
|/ | | | uniforms/attributes); Refine API doc.
* NEWT-MouseEvent getWheelRotation() API Update - Fixes Bug 659: NEWT ↵Sven Gothel2013-01-1422-250/+724
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Horizontal Scrolling Behavior (OSX, X11, Win32); Bug 639: High-Res Mouse-Wheel - API update 'float getWheelRotation()': Usually a wheel rotation of > 0.0f is up, and < 0.0f is down. Usually a wheel rotations is considered a vertical scroll. If isShiftDown(), a wheel rotations is considered a horizontal scroll, where shift-up = left = > 0.0f, and shift-down = right = < 0.0f. However, on some OS this might be flipped due to the OS default behavior. The latter is true for OS X 10.7 (Lion) for example. The events will be send usually in steps of one, ie. -1.0f and 1.0f. Higher values may result due to fast scrolling. Fractional values may result due to slow scrolling with high resolution devices. The button number refers to the wheel number. - Fix Bug 659: NEWT Horizontal Scrolling Behavior (OSX, X11, Win32) - See new API doc above - X11/Horiz: Keep using button1 and set SHIFT modifier - OSX/Horiz: - PAD: Use highes absolute scrolling value (Axis1/Axis2) and set SHIFT modifier for horizontal scrolling (Axis2) - XXX: Use deltaX for horizontal scrolling, detected by SHIFT modifier. (traditional) - Windows/Horiz: - Add WM_MOUSEHWHEEL support (-> set SHIFT modifier), but it's rarely impl. for trackpads! - Add exp. WM_HSCROLL, but it will only be delivered if windows has WS_HSCROLL, hence dead code! - Android: - Add ACTION_SCROLL (API Level 12), only used if layout is a scroll layout - Using GestureDetector to detect scroll even w/ pointerCount > 2, while: - skipping 1st scroll event (value too high) - skipping other events while in-scroll mode - waiting until all pointers were released before cont. normally - using View config's 1/touchSlope as scale factor - Fix Bug 639: High-Res Mouse-Wheel - getWheelRotation() return value changed: int -> float allowing fractions, see API doc changes above. - Fractions are currently supported natively (API) on - Windows - OSX - Android - AndroidNewtEventFactory ir refactored (requires an instance now) and AndroidNewtEventTranslator (event listener) is pulled our of Android WindowDriver.
* Fix Bug 666: GLBase.java line 438 contained some invalid trailing characters ↵Sven Gothel2013-01-121-1/+1
| | | | (regression from e7064ece049705e013d80985eae698ce0ee3c4e3)
* Move all Exclusive Context Thread (ECT) tests to it's own package to ↵Sven Gothel2013-01-1212-12/+12
| | | | de-clutter core test package
* Adding GEOMETRY_SHADER support in ShaderCode, adding core ↵Sven Gothel2013-01-1212-52/+610
| | | | | | | | | | | | | | | | | | | GL3/GEOMETRY_SHADER unit tests. ; Simplified GLContext version number - Adding GEOMETRY_SHADER support in ShaderCode, adding core GL3/GEOMETRY_SHADER unit tests Chuck Ritola reported in December 2012 that we lack support of GEOMETRY_SHADER and he provided a test case. The latter is cleaned up to use GL3 core profile features only tesing a pass-through and the flip-XYZ geometry shader. ShaderUtil is fixed. - Simplified GLContext version number The OpenGL major/minor version is now hold in a VersionNumber instance to simplify usage. Also expose it via getGLVersionNumber() while marking getGLVersionMajor() and getGLVersionMinor() deprecated.
* Add old Olympic demo from GL4Java - Make it use ES1 and suitable for ↵Sven Gothel2013-01-112-0/+498
| | | | FixedFunctionEmul.
* Minor: GLWindow fix size validation; GLContextImpl: Remove hold ctx lock ↵Sven Gothel2013-01-112-7/+4
| | | | count constraints at destroy.
* GLAutoDrawable/AnimatorBase: Add ExclusiveContextThread (ECT) feature; ↵Sven Gothel2013-01-1130-470/+2782
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | AnimatorBase: Add setModeBits/MODE_EXPECT_AWT_RENDERING_THREAD; FPSAnimator: Make transactions deterministic. ExclusiveContextThread (ECT) allows user to dedicate a GLContext to a given thread. Only the ECT will be allowed to claim the GLContext, hence releasing must be done on the ECT itself. The core feature is accessible via GLAutoDrawable, while it can be conveniently enabled and disabled via an AnimatorBase implementation. The latter ensures it's being released on the ECT and waits for the result. Note that ECT cannot be guaranteed to work correctly w/ native (heavyweight) AWT components due to resource locking and AWT-EDT access. This is disabled in all new tests per default and noted on the API doc. Note: 'Animator transaction' == start(), stop(), pause(), resume(). - Add ExclusiveContextThread (ECT) feature - GLAutoDrawable NEW: - Thread setExclusiveContextThread(Thread t) - Thread getExclusiveContextThread() - AnimatorBase NEW: - Thread setExclusiveContext(Thread t) - boolean setExclusiveContext(boolean enable) - boolean isExclusiveContextEnabled() - Thread getExclusiveContextThread() - AnimatorBase: Add setModeBits/MODE_EXPECT_AWT_RENDERING_THREAD Allows user to pre-determine whether AWT rendering is expected before starting the animator. If AWT is excluded, a more simple and transaction correct impl. will be used. - FPSAnimator: Make transactions deterministic. FPSAnimator previously did not ensure whether a transaction was completed. A deterministic transaction is required to utilize ECT. FPSAnimator now uses same mechanism like Animator to ensure completeness, i.e. Condition and 'finishLifecycleAction(..)'. Both are moved to AnimatorBase. Tested manually on Linux/NV, Linux/AMD, Windows/NV and OSX/NV. - All new tests validated correctness. - All new tests shows an performance increase of ~3x w/ single GLWindow, where multiple GLWindows don't show a perf. increase.
* Fix GLArrayDataClient createGLSL(..) 'name' setting; Also adds equal ↵Sven Gothel2013-01-031-3/+3
| | | | | | | | | behavior for Client/Server FFP usage. - GLArrayDataWrapper's init() only validated and set the 'name' for VBO ARRAY_BUFFER usage and validates it, regardless of FFP or GLSL usage. - GLArrayDataWrapper's init() included the no-VBO case to the above 'name' set/validate code path.
* Update PNGJ 0.85 -> 1.12 (w/ interlace read support) ; Added PNG Interlace ↵Sven Gothel2012-12-3155-1454/+3943
| | | | read tests (TestPNGTextureFromFileNEWT)
* Enhance PMVMatrix: Use FloatStack to implement PushMatrix and PopMatrix (Bug ↵Sven Gothel2012-12-311-29/+28
| | | | | | | | | | | | | 657) One FloatStack for each, MV, P and T, with initialialSize zero to save memore and growSizes: MV = 16 arrays, P = 2 arrays, T = 2 arrays, with array = 16 floats. This shall save performance due to the preallocated stack when used and growing only in the above mentioned grow intervals. We may finetune the growSize, if required.
* NEWT WindowsWindow.c: UpdateInsets: Fix determiniation of isUndecorated - ↵Sven Gothel2012-12-301-1/+1
| | | | WS_SYSMENU is _not_ an indication!
* Bug632: Test NEWT Child Window Translucency (X11/Windows) ..Sven Gothel2012-12-306-15/+180
| | | | | | | | | | | | | - Windows: Child window is not translucent at all - X11: Child window is translucent to parent's background, however - parents content is _not_ 'composed in'. - TODO: Find whether there is a solution or not. - Note: The child window does not change it's rel. position if parent moves! This is a feature, since we don't have impl. a layout.
* SharedResourceRunner: Always dump Exception if catched @ creation, not just ↵Sven Gothel2012-12-301-3/+1
| | | | | | in DEBUG case. This seems to be crucial, since otherwise user is left in the dark w/o enabling debug flags.
* setGLFunctionAvailability(..): Unify GLInt and GLString version validation; ↵Sven Gothel2012-12-301-50/+102
| | | | | | | | | | | | | | | | | Refines commit b6ae4e4dcbd740dd57de9dc3280d943e98cdaa76 - Limit getGLIntVersion() w/ non ARB ctx to 3.0 - similar to getGLVersionNumber(.., String). - GLInt only proceeds w/ valid version - Always test GLInt version before GLString version, cont. w/ GLString version if GLInt failed. - Fail if strictMode and GLInt failed, but GLString reports GL >= 3.0 - Fail is strictMode and neither GLInt nor GLString could validate version - Clear glError before succeeding (could be tainted by GLInt version test)
* Cleanup / Simplify: setGLFunctionAvailability(..) / ↵Sven Gothel2012-12-302-50/+117
| | | | | | | | | | createContextARBVersions(..) GL Version Validation String or integer based GL version validation now happens in setGLFunctionAvailability(..) depending on the requested profile. Due to the 'strictMatch' argument the method fails early when unsatisfied also allowing to simplify createContextARBVersions(..) implementation.
* Fix Bug 658 (Mesa 9.0 3.1 Intel compat quirk, 3.1 core only) ; No ↵Sven Gothel2012-12-294-43/+306
| | | | | | | | | | | | | | | | | | | PROFILE_ALIASING compat -> core ; Fix setGLFunctionAvailability(..) failure path @ profile query - Add GLRendererQuirks.GLNonCompliant, marking a GL context/profile non compliant. Currently: 'Mesa DRI Intel(R) Sandybridge Desktop' && 3.1 compat profile - Fix Bug 658 (Mesa 9.0 3.1 Intel compat quirk, 3.1 core only) Detect case using new GLRendererQuirks.GLNonCompliant in setGLFunctionAvailability() and return 'false'. - No PROFILE_ALIASING compat -> core Use true core GL profiles / context if available to ensure proper API behavior across platforms due to different functionality. E.g. don't use GL3bc if GL3 is requested. - Fix setGLFunctionAvailability(..) failure path @ profile query Destroy temp context & zero result to cont. iterating through GL versions. This missing cleanup lead to returning the faulty GL context handle and it's mapping/usage.
* Adapt change of commit d93c5d23e304ea20e868595748f92a5bef4f5703 to unit test ↵Sven Gothel2012-12-291-26/+76
| | | | (GLCanvas realization point)
* AWT GLCanvas: More strict GLDrawable realization [on AWT-EDT], skip if ↵Sven Gothel2012-12-282-7/+8
| | | | | | | | | | | | | | | | | | | | creation is not possible on AWT-EDT. The Intel HD3000 OpenGL driver on Windows will deadlock @ SwapBuffers in case the drawable is created on a thread other than the window owner thread. We are aware of such possibilities, nevertheless the AWTEDTExecutor.singleton.invoke(..) allowed to execute the runnable in case it cannot be invoked on AWT-EDT. The latter is the case if the current thread is not the AWT-EDT _and_ is holding the AWT tree-lock. With GlueGen commit 0b43b43f889ad7fc220942b0076e2001ca3cf13f, the invoke method now consumes an argument allowing to restrict the execution to AWT-EDT only. In such case, the drawable will be realized at a later time from the AWT-EDT. Such a situation could be triggered if a Frame's setVisible(true) is not issued from the AWT-EDT, as it should be! However, to relax such use cases - we better recognize such possible dealock and avoid it.
* NEWT/Android: Fix NewtBaseActivity.getWindow() recursion if used w/o ↵Sven Gothel2012-12-251-1/+1
| | | | delegated Activity, i.e. our ActivityLauncher (Completes commit a35beb22d712b6da85a794115b19d484a12c8643)
* Test*NewtEventModifiers*: Clear 'clearing' events on thread - @Before/@After ↵Sven Gothel2012-12-251-42/+32
| | | | | | -> execOffThreadWithOnThreadEventDispatch Remaining 'clearing' events from _releaseModifiers() and escape() broke test case.
* Misc OSX/SWT: OSXUtil.RunOnMainThread(..) refinement; Fix ↵Sven Gothel2012-12-256-91/+205
| | | | | | | | | | | | | | | | | Test*NewtEventModifiers for SWT (TestNewtEventModifiersNewtCanvasSWT) - Misc OSX/SWT: OSXUtil.RunOnMainThread(..) refinement - 'waitUntilDone' is implemented on Java site via lock/wait on RunnableTask to not freeze OSX main thread. - Fix Test*NewtEventModifiers for SWT (TestNewtEventModifiersNewtCanvasSWT) - Deal with SWT's requirement to run the SWT event dispatch on the TK thread, which must be the main thread on OSX. We spawn off the actual test-action into another thread, while dispatching the events until the test-action is completed. - AWTRobot: Add 'void requestFocus(Robot robot, Object obj, int x, int y)' - Use waitForIdle() only if programmed in Robot (Deadlock w/ OSX SWT) - Required for SWT usage (see above)
* NEWT/Android: Fix NewtBaseActivity.getWindow() recursion if used w/o ↵Sven Gothel2012-12-241-2/+6
| | | | delegated Activity, i.e. our ActivityLauncher
* WT-NEWT Modifier mapping and test: part-4 (Fix unit tests): ↵Sven Gothel2012-12-242-75/+117
| | | | BaseNewtEventModifiers ignore more events (clicked, wheel, entered, exited); Fix TestNewtEventModifiersNewtCanvasSWT SWT-Display thread.
* NEWT peserve 'mouseButtonModMask' to be sent for all mouse- and key events. ↵Sven Gothel2012-12-241-30/+43
| | | | | | | | | | | | | | | | TODO: NEWT Event Factories. Misc: Cleaned up spacing. On some native OS, the accumulation of pressed button modifier-mask is not available, e.g. OS X. NEWT WindowImpl.doMouseEvent(..), invoked by native NEWT events, will track the pressed mouse button modifier-mask, similar to mouseButtonPressed to synthesize the DRAGGED event. Added NEWT WindowImpl.doKeyEvent(..) to honor the pressed mouse button modifier-mask, i.e. pass it w/ key events as well. TODO: Unify synthesization of NEWT event artifacts as described by the above, allowing NEWT event translation to benefit from same code to gain same semantics. Notable: AWTNewtEventFactory and SWTNewtEventFactory
* AWT-NEWT Modifier mapping and test: part-3 (Fix unit tests)Sven Gothel2012-12-245-167/+208
| | | | | | | | | | | | | | | | | | | | | | | - Finetune delay - Wait for eventCount - Fix Listener concurrency - Manually tested (enabled in our unit tests now) - Tool Combinations - NEWT GLWindow - AWT GLCanvas - NewtCanvasAWT - NewtCanvasSWT - On - Linux/X11 - Windows - OSX(+) (+): Failure NEWT: When multiple buttons are pressed, only the last one is visible via modifier MASK. "expected:[button1, button2], have: [button2]"
* AWT-NEWT Modifier mapping and test: part-3 (NEWT BUTTON MASK always, ..)Sven Gothel2012-12-246-125/+135
| | | | | | | | | | | | | | | | - AWTNewtEventFactory's awtModifiers2Newt: - always include NEWT BUTTON_MASK (press, release, ..) where AWT doesn't include them at release (it's only a DOWN_MASK). - Test BaseNewtEventModifiers, .. - No need to call super class Before, BeforeClass, .. manually - Use RedSquareES2 as GL demo - Adapt to AWTNewtEventFactory's modifier change above (NEWT BUTTON MASK even at release) - More descriptive error/log text - Added _mandatory_ TestNewtEventModifiersNEWTWindowAWT to test native NEWT behavior. This shall be the golden behavior all translated events shall compare w/.
* Move pre-existing NEWT event test to new dedicated package package: ↵Sven Gothel2012-12-236-12/+5
| | | | com.jogamp.opengl.test.junit.newt.event
* Fix 13168c99ff9e8bf71c83f1be7afee270a3db4074 / ↵Sven Gothel2012-12-237-173/+164
| | | | | | | | | | | | | | | | | 811e3791b98fea0dfa3b7d301cb532c54df8dc82: AWT-NEWT Modifier mapping - part-2 AWTNewtEventFactory: - getAWTButtonMask() -> getAWTButtonDownMask() - using proper _DOWN_MASK values (regression of commit 13168c99ff9e8bf71c83f1be7afee270a3db4074) - com.jogamp.newt.event.MouseEvent.BUTTON_NUMBER buttons - adding 'ModifierMappings.txt' to API doc header - remove obsolete 'int awtModifiers2Newt(int awtMods, boolean mouseHint)' - 'int awtButton2Newt(int awtButton)' 1:1 button name mapping Tests: - rename TestNewtEventModifiers -> BaseNewtEventModifiers to avoid being picked up by our junit testing framework. The latter tests all classes starting w/ 'Test*'
* Fix commit 811e3791b98fea0dfa3b7d301cb532c54df8dc82: Make AWT usage Java6 ↵Sven Gothel2012-12-232-63/+66
| | | | clean (was using Java7 stuff); Note: Need to test!
* Merge pull request #54 from rhatcher/masterSven Gothel2012-12-236-2/+1258
|\ | | | | Fix For Bug 629, Plus First Crack At Associated Unit Tests
| * Merge branch 'master' of https://github.com/sgothel/joglrhatcher2012-12-0621-484/+1632
| |\
| * | Fix for JOGL bug 629, and added new unit tests.rhatcher2012-11-296-2/+1258
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The change to AWTNewtEventFactory appears to fix the original issue for AWTCanvas instances, and the TestNewtEventModifiersAWTCanvas appears to work ok too. However, there are still issues with NewtCanvasAWT and NewtCanvasSWT instances. These might be problems in the test code, but there's also a good chance there are still issues in the NEWT event delivery infrastructure. For the time being I recommend that only TestNewtEventModifiersAWTCanvas be included in routine unit tests. The tests are defined in TestNewtEventModifiers, and the remaining test classes extend it to define how the window and associated GL drawing surface are created. File ModifierMappings.txt is simply informational, and shows how the modifier bits are laid out between AWT and NEWT. This possibly should have been a spreadsheet.
* | | Bug 642 / Refine 1ae0737f34143a5ed655bd9c4d5fe9b0437c7774: ↵Sven Gothel2012-12-221-2/+2
| | | | | | | | | | | | | | | | | | | | | GLCanvas.displayOnEDT: Also check null!=drawable As for 1ae0737f34143a5ed655bd9c4d5fe9b0437c7774, an animator may inject a display Runnable on the EDT before AWT destruction. In case this Runnable is executed after destruction on the EDT - it would fail.
* | | Fix regression of commit b8a8fc24a3afb0cb06a31504bdea1a98b8f00ef4: Cache ↵Sven Gothel2012-12-221-4/+5
| | | | | | | | | | | | ShaderState for share ctor.
* | | Bug 642 / Refine 1ae0737f34143a5ed655bd9c4d5fe9b0437c7774: ↵Sven Gothel2012-12-221-48/+42
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | GLContextImpl.makeCurrent() - Handle !drawable.isRealized() early, don't catch 'create' Exception Catching the createImpl() exception could be confusing, since it shall succeed. Such exception is required to hint a platform bug and to debug it. Hence testing drawable.isRealized() upfront is preferrable, i.e. catching a well known case for returning CONTEXT_NOT_CURRENT.
* | | Bug 642: Refine test case, adding Hw/Lw mix intermediate Container test to ↵Sven Gothel2012-12-222-8/+24
| | | | | | | | | | | | avoid remove/add of GL component on Window when moving splitter.
* | | Fix Bug 642 TestJSplitPaneMixHwLw01AWT (AWT-GLCanvas); Robustness ↵Sven Gothel2012-12-226-27/+144
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | GLContext/GLDrawable - Fix Bug 642 TestJSplitPaneMixHwLw01AWT On Windows platform when mixing hw/lw JSplitPanel, the GLCanvas is removed and added when splitter is moved. The lack of robustness (see below) lead to an exception. Note: Only w/ GLJPanel (no hw/lw mixing) the splitter can be moved in both direction. Only here it is guaranteed that the GL component will survive the action. - Fix AWT-GLCanvas EDT Runnable: swapBuffer().. / display(..) - Check drawable.isRealized() within the lock on the performing thread. This is not possible before issuing the EDT Runnable action since we cannot hold the lock beforehand. - Robustness GLDrawableImpl - boolean realized -> volatile boolean realized - remove 'synchronized' on isRealized() and setRealized(..) - Use dbl-checked locking on 'realized' test for swapBuffers() and setRealized(..) - Robustness GLContextImpl - Catch createImpl(..) exception and properly return CONTEXT_NOT_CURRENT
* | | *Drawable impl. DEBUG: Add getThreadName() to debug out.Sven Gothel2012-12-228-13/+16
| | |
* | | FBObject, minor edit: GL2GL3.GL_MAX_COLOR_ATTACHMENTS -> ↵Sven Gothel2012-12-211-1/+2
| | | | | | | | | | | | GL2ES2.GL_MAX_COLOR_ATTACHMENTS
* | | GLContextImpl: Make createContextARBImpl/setGLFunctionAvailability more ↵Sven Gothel2012-12-219-23/+32
| | | | | | | | | | | | robost while detecting erroneous queried GL version
* | | Bug 651: Mesa3D: GL_INVALID_ENUM in ↵Sven Gothel2012-12-202-2/+212
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | glGetIntegerv(pname=GL_MAX_TEXTURE_IMAGE_UNITS) ; Add unit test / bisect Mesa3D This bug lies within Mesa3D (any renderer) and is fixed in commit 8dc79ae7d73cf6711c2182ff9a5d37ef6c989d23. Mesa3D Version 9.0 still exposes this bug, where 9.0.1 has it fixed w/ above commit.
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-15 22:49:04 +0000