This crash was discovered by Dalai and this happened because of
unset current context (as result of call wglMakeCurrent(NULL, NULL)).
In this case glGetString(GL_VENDOR) returns NULL. Rather than add check
for vendor != NULL before string comparison, I've changed a bit logic of
context creation:
- Create context and set it as current
- If it's crappy Intel card -- delete this context and
share the only one context between all Windows
- Otherwise, use initial logic (with sharing lists and so on)
This could also fix crash when opening userprefs from a menu with Intel card.
* memory corruption when skipping over long marker (was attempting to read over end of file)
* also updated internal jpeg macros to be the same as in jpeg lib
Plugged the eltopo library into the cloth solver.
I was playing with it earlier, and it's so easy to
use I decided to quickly put it in (trunk's) cloth.
See http://www.cs.ubc.ca/labs/imager/tr/2009/eltopo/eltopo.html
. The authors are on the bleeding edge of continuous collision
detection (one of them did ILM's cloth sim).
I
don't really have to time to plug it into softbody, particles,
bullet, fluid, etc, but doing so would be pretty straightforward.
I'll leave that up to someone else.
To use, turn on USE_ELTOPO (in cmake) or WITH_BF_ELTOPO in scons.
Intel video cards don't work fine with multiple contexts and
have to share the same context for all windows. This could work
incorrect with multiple video cards configuration, so suppose
there'll be no such situation (Intel cards are mostly in portable
devices like notebooks and laptops, where there's actually no
dual video cards). Anyway, it should work much better now.
Non-Intel cards behavior was kept unchanged.
Thanks to Ton for debug session :)
if not defined (first run) these are now set blank but can be defined later.
the problem is that scons & cmake builds would link against different libraries since cmake added its own defaults.
now, by default, scons & cmake have the same libraries.
This fixes an obscure crash in MinGW where cmakes default linking with -ladvapi32 would crash on string formatting which used float precision as an argument, eg:
printf("%.*f", 3, value);
...without giving a useful backtrace or pointing to the line of code doing the string formatting.
- Assigning local materials to library objects disabled
(crashes on undo/redo cases)
- Disabling options in Material buttons to add/remove slots
on library data
- Drawing Object ID template in Object properties, this
allows browse active Object, but especially shows library
status then.
submitted by Tom Edwards
This patch introduces a switch -con and its longer version --start-console. When giving this on cmd-line you'll get the black console window. The new behaviour is to hide it by default. We'll still see briefly the console at startup and during exit, but that's something that cannot be changed.
If you start blender from a cmd.exe, the console will not be hidden.
was no longer wrapped by python as a vector. now fixed size float arrays
with PROP_NONE subtype are wrapped as vectors since its convenient to
have x/y access.
On anim-render, a click in timeline stopped render completely.
The reason for this was a bit wacko code to cope with frame-step
feature (steps of multiple frames).
I thought of fixing that, but instead decided to block any operator
in Blender to change a frame while a render is in progress.
Both render engine and UI are accessing (writing to) the same
data then, which is a bad conflict.
Still a serious weakness of threaded render, but I'll keep
trying to allow this as far as possible :)
