Hi All,
This is a 'fix' to sgscreen library,
for those with a glX extension implementation
that still reports version 1.2...
Wow, I got there!!! Boy I love coding when
you persist, and try, and try some more,
and you discover what makes it work ;=))
even when you do not fully 'understand'
all that is happening...
The discovery :
(a) While my installed implementation of
the service :-
glXQueryVersion(dpy, &major, &minor)
returns 1.2, it none the less supports
many features, services of 1.3!
(b) While my installed implementation of -
const char* extString =
glXQueryExtensionsString(dpy, screen);
does NOT contain the string "GLX_SGIX_pbuffer",
BUT a call to the service :-
const char * extClient =
glXGetClientString(dpy,GLX_EXTENSIONS);
DOES...
This can be seen when running 'glxinfo'... so I added
this and continued...
(c) Although the documentation for the service
glXCreateGLXPbuffer seems to indicate that a NULL
attribute pointer can be used, this produced a
segfault in my system, so I add a pointer to
attributes (which had already been declared,
except I had to move it to a higher context)...
But in my case it still reported an X error, but
at least not a segfault, so...
(d) Since my implementation returns 1.2, the code logic of -
if (((1 <= major && 3 <= minor) || 2 <= major) && ...
naturally stops 'glXVersion1_3Present' being set TRUE,
even though ALL the 1.3+ functions are resolved
using SGLookupFunction("glXChooseFBConfig"), etc
And the code then seeks 'alternate' functions, like
SGLookupFunction("glXChooseFBConfigSGIX");
which are also _ALL_ found, but yield 'bad' results
when used...
So in 2 places in the code, I overrode the simple
if ( glXVersion1_3Present ), which is false in
my case, with more logic :-
Reasoning: If the 1.3 pointers are resolved, then
use them, instead of the old, failing!, 1.2
functions...
1.// if (glXVersion1_3Present)
if (glXCreatePbufferPtr && glXGetVisualFromFBConfigPtr &&
glXCreateContextPtr)
{
2.// if (!glXVersion1_3Present)
if ((!glXCreatePbufferPtr || !glXGetVisualFromFBConfigPtr || !
glXCreateContextPtr) &&
(!glXVersion1_3Present))
{
BUT, this did NOT work in a 3rd case I tried :
---------------------------------------------
// if (glXVersion1_3Present)
if ((glXVersion1_3Present) ||
(glXCreatePbufferPtr && glXGetVisualFromFBConfigPtr &&
glXCreateContextPtr && glXQueryDrawablePtr))
so left this back the way it was... the
GLXDrawable draw = glXGetCurrentDrawable();
worked, but the glXQueryDrawablePtr for GLX_WIDTH and GLX_HEIGHT
FAILED, and exited, with an X error.
AND glXQueryGLXPbufferSGIXPtr(...) with GLX_WIDTH_SGIX
and GLX_HEIGHT_SGIX also FAILED with -
X Error of failed request: GLXBadPixmap
This seems troublesome in that _iWidth and
_iHeight are left as per the buffer
request, and not adjusted if the buffer
size is different to that requested.
But, on checking further, at the time
this is done, the context and buffer
has been returned to the original, so I
see no reason, or problem, that this code
has been 'chopped' ;=()
-------------------------------------------
*** ===========================
*** THEN IT ALL WORKED ;=))
*** ===========================
When I run TestRenderTexture I have a spinning
rectangle, containing a spinning torus, and
can toggle each of the init strings
(with the enter key) - most work, but some do
not, but I suspect this is true in ALL cases,
environments.
And the space bar toggles the color/depth
and effects some images, but not all...
which I think is correct.
The FIX :
--------
To get here I added over 100 dbg_printf
statements, and hope that these are kept
to help future debugging. They can ALL
be defined away with -
#define dbg_printf
or enabled anytime with -
#define dbg_print printf
Some of these are in addition to an
existing SG_LOG(), but since the
application TestRenderTexture does NOT
'enable' sglog(), then none of these
show... hence the simple 'dbg_printf'
alternative added...
And some other debug code depends on
_DEBUG being defined... This includes a
big block of DEBUG code added to the
end of the file.
Version:
Also, for no particular reason I did
all these changes first on SimGear-1.9.0,
but was pleased to note that there are NO
CVS changes to the files changed, so
the attached 'patch' can be applied
directly to SG CVS...
Which is what I have done to my local
cvs copy, and the diff attached is using -
...simgear/source $ cvs diff -u
If someone would like to check my work,
I would be pleased to answer any question
as to why this, or that...
And of course, those that had the good
luck to have the TestRenderTexture working
in their system need to re-check that
nothing I did effects it still working in
their system... more reason to KEEP
the copious 'debug' code in place...
In most places, I only touched code in
the NOT _WIN32, NOT __MACH__ sections,
so only *nix code should be effected.
Use:
Of course, I have tested the new sgscreen
static library with Atlas/Map, and now
my Map also works flawlessly... ;=))
So the ONLY remaining thing is to get
these changes into SG CVS so the 'torus'
can continue to spin ;=))
Regards,
Geoff.
attached: oops, not attached - too big 40KB++
so only available -
online: http://geoffair.net/tmp/sg-rt-diff.patch
------------------------------------------------------------------------------
_______________________________________________
Flightgear-devel mailing list
Flightgear-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/flightgear-devel
