Age | Commit message (Collapse) | Author | Files | Lines |
|
This allows the user to continue to type at other windows while glean
tests are running.
|
|
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
|
|
Also add tinderbox install root to include and library paths
Also ensure $(DESTDIR)/$(prefix) directory exists after install, otherwise jhbuild complains
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
|
|
MultiTest::compareOne neglects to report the name of the test
which it is comparing the results for
Some tests report just "New: FAIL Old: FAIL", omitting the test
name and without a trailing '\n', so they run into the next test's
results (which is irritating). On further examination, they are using
comparePassFail() which would report this information in any case
if it was interesting (because the result had changed).
GetStringTest::compareOne doesn't emit a newline after reporting the
differing strings, and so runs into the output of the next test
|
|
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
|
|
Therefore avoiding to have to install the MSVC runtime redistributables.
|
|
|
|
Include is necessary when building with WIN32_LEAN_AND_MEAN.
|
|
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
Prevents conflicting declaration errors inside glext.h with recent
MinGW compiler toolchains.
|
|
GLEAN::Window::~Window was calling XDestroyWindow on the currently
bound GLX drawable, which would cause the subsequent test to explode
when it called glXMakeCurrent and Mesa attempted to flush the
outstanding state to the now destroyed window.
Now the Window destructor checks if the window being destroyed is
bound to the current GLX context, and if so, releases the current GLX
context before destroying the window.
I also included the equivalent WGL change, but that is untested.
The AGL and BeOS versions are untouched, but probably needs something
similar.
With this change, the glean basic sanity test now passes on Mesa swrast.
Signed-off-by: Nicholas Miell <nmiell@gmail.com>
Signed-off-by: Brian Paul <brianp@vmware.com>
|
|
Signed-off-by: José Fonseca <jfonseca@vmware.com>
|
|
This patch fixes several version checks that are too specific
(e.g. checking that the GL version is 2.x or 3.x when the intent is to
make sure the GL version is >= 2.0).
We now perform the version check using GLUtils::getVersion() function.
Signed-off-by: Brian Paul <brianp@vmware.com>
|
|
The values were being interpreted as chars before.
|
|
MinGW doesn't understand it and will produce warnings.
|
|
Creates havoc when linking against certain libraries.
This reverts commit 81469c15be2a5abc49616fd6e51b74dba2b1fbb7.
|
|
|
|
|
|
dlsym() is all one needs.
|
|
Avoids having to install the MSVC redistributable runtime.
|
|
|
|
|
|
Broken on windows 64 bits, and obsoleted by CMake, which needs much
less customization/tweaking to work as desired.
|
|
Tested with Linux GCC, MSVC, and MinGW crosscompilers.
|
|
cast to double added to the first argument expression to prevent the integer overflow
Signed-off-by: Simone Contini <s.contini@oltrelinux.com>
Signed-off-by: Brian Paul <brianp@vmware.com>
|
|
|
|
Unnecessary/undocumented.
|
|
|
|
|
|
Using the same approach as texCombine.
Logged all combinations and ensured that the frequencies match the desired.
|
|
constructor.
|
|
|
|
|
|
|
|
Fixes build with GCC 4.4.x.
|
|
|
|
|
|
cstdio was removed as part of previous commit
53d45c4d7a4db12c90a34fbb0cedd1dd2a79a02d but is needed for snprintf.
|
|
|
|
glIsBuffer() should return GL_FALSE if the buffer was generated with
glGenBuffers() but never bound with glBindBuffer().
This was incorrect in Mesa until yesteday. Mesa now agrees with
AMD, Apple and Intel drivers. NVIDIA is the odd one out.
Also, improve error messages and remove another unneeded glIsBuffer()
call later.
|
|
Fixes GL_INVALID_ENUM errors on Mac OS X.
|
|
Fixes a valgrind uninitialized value error.
|
|
"Fixes" the HTML output from piglit-summary-html.py.
|
|
This compensates for slight differences in line rasterization from one
driver to another and allows more drivers to pass.
|
|
GL lets the implementation be +/- a pixel on line location. It was
getting rounded down on the intel driver, so the lines weren't drawn
to 1D FBOs. This should have a better chance on drivers in general.
(cherry-picked from piglit commit 9fdd8d9d1fc56178c950ba8fbca8abfb2e9d8b96)
|
|
Fixes spurious test failures on the intel driver.
(cherry-picked from piglit 35122d69307f0b9c88bc143715f616be80ac5cae)
|
|
Before, we read one row of pixels where we expected the point to be
in order to compute the point size. With some drivers single pixel
points might be positioned one above or below the expected position
thus causing the test to fail. Now, read thee rows of pixels to make
sure we don't miss the point. The purpose of this test is to measure
attenuated point size, not exact positioning of single pixel points.
|
|
Redeclare the gl_TexCoord[] array with a specific size, per the GLSL spec.
Note that the GLSL Orange book says that gl_TexCoord[] is declared with
gl_MaxTextureCoords size while the GLSL 1.x specs declare it unsized.
|
|
|
|
|