On 03/01/2012 11:39 AM, Benjamin Franzke wrote:
> 2012/3/1 Chad Versace <chad.vers...@linux.intel.com>:
>> On 02/29/2012 07:36 AM, Benjamin Franzke wrote:
>>> We dont want eglplatform.h to typedef egl native types
>>> to x11 types, when x11 headers are not available.
>>> ---
>>> configure.ac | 4 ++++
>>> 1 files changed, 4 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/configure.ac b/configure.ac
>>> index 0caa1b1..92a0e52 100644
>>> --- a/configure.ac
>>> +++ b/configure.ac
>>> @@ -1402,6 +1402,10 @@ if test "x$enable_egl" = xyes; then
>>>
>>> AC_CHECK_FUNC(mincore, [DEFINES="$DEFINES -DHAVE_MINCORE"])
>>>
>>> + if test "$have_x" != yes; then
>>> + DEFINES="$DEFINES -DMESA_EGL_NO_X11_HEADERS"
>>> + fi
>>> +
>>> if test "$enable_static" != yes; then
>>> # build egl_glx when libGL is built
>>> if test "x$enable_glx" = xyes; then
>>
>> After examining the where the macro is used, in eglplatform.h ...
>>
>> --- snip ---
>> #elif defined(__unix__)
>>
>> #ifdef MESA_EGL_NO_X11_HEADERS
>>
>> typedef void *EGLNativeDisplayType;
>> typedef khronos_uint32_t EGLNativePixmapType;
>> typedef khronos_uint32_t EGLNativeWindowType;
>>
>> #else
>>
>> /* X11 (tentative) */
>> #include <X11/Xlib.h>
>> #include <X11/Xutil.h>
>>
>> typedef Display *EGLNativeDisplayType;
>> typedef Pixmap EGLNativePixmapType;
>> typedef Window EGLNativeWindowType;
>>
>> #endif /* MESA_EGL_NO_X11_HEADERS */
>> --- end snip ---
>>
>> I can't think of a valid reason for the #else branch to ever be taken.
>> If you're building EGL on Unix/Linux, but not for gbm (__GBM__), Android
>> (ANDRDOID),
>> Wayland (WL_EGL_PLATFORM), or X (the default), then for what platform
>> are you building libEGL? (The question is not rhetorical).
>
> The platform headers for wayland/gbm are not included when compiling
> src/egl/main/* or some other common code in st/egl, that is when the
> else branch is taken.
> So currently we include Xlib.h when building without x11 platform,
> that fails if its not available.
> Will add a note to the commit message.
It appears that the patch may introduce a bug in the 64-bit build of libEGL,
but I'm not certain. Here's my reasoning.
1. Suppose that
- This patch is applied.
- The system has no X headers installed.
- Some Wayland application has a source file, broken.c, that
#includes <wayland-egl.h> then <EGL/egl.h>.
2. In eglapi.c, which defines eglCreateWindowSurface, EGLNativeWindowType
will be defined as khronos_uint32_t.
3. In broken.c, wayland-egl.h first defines WL_EGL_PLATFORM. Then
EGLNativeWindowType will be defined as wl_egl_window*, a 64-bit type.
Without the patch, eglapi.c safely typedefs EGLNativeWindowType to XID, which
matches
the system's pointer size.
- Chad
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
