Viewer LL 3.3.4 and before sometimes fail to properly redisplay dynamic textures that have a small data length compared to pixel size when pulled from cache.
This appears to happen when the data length is smaller than the estimate discard level 2 size the viewer uses when making this GetTexture request.
This commit works around this by always regenerating dynamic textures that fall below this threshold rather than reusing them if ReuseDynamicTextures = true
This can be controlled by the [Textures] ReuseDynamicLowDataTextures config setting which defaults to false.
This is because recent viewers (3.2.1, 3.3.4) and probably earlier ones using the http GetTexture capability will sometimes make such invalid range requests.
This appears to happen if the viewer's estimate of texture sizes at discard levels > 0 (chiefly 2) exceeds the total texture size.
I believe this does not normally happen but can occur for dynamic textures with are large but mainly blank.
If this happens, returning a RequestedRangeNotSatisfiable will cause the viewer to not render the texture at the final resolution.
However, returning a PartialContent (or OK) even with 0 data will allow the viewer to render the final texture.
Disabled (status quo) by default.
This flag makes the dynamic texture module reuse cache previously dynamically generated textures given the same input commands and extra params for 24 hours.
This occurs as long as those commands would always generate the same texture (e.g. they do not contain commands to fetch data from the web).
This makes texture changing faster as a viewer-cached texture uuid is sent and may reduce simulator load in regions with generation of lots of dynamic textures.
A downside is that this stops expiry of old temporary dynamic textures from the cache,
Another downside is that a jpeg2000 generation that partially failed is currently not regenerated until restart or after 24 hours.
In testing, it appears that if multiple threads dispose of separate GDI+ objects simultaneously,
the native malloc heap can become corrupted, possibly due to a double free(). This may be due to
bugs in the underlying libcairo used by mono's libgdiplus.dll on Linux/OSX. These problems were
seen with both libcario 1.10.2-6.1ubuntu3 and 1.8.10-2ubuntu1. They go away if disposal is perfomed
under lock.
The convention is that if an object implements IDiposable() the code must explicitly call Dispose() or call it via the using statement.
This may be particularly important for GDI+ objects since they encapsulate native code entities.
Added the following replacement functions for compliance to the OSSL standards stated on the wiki:
osGetTerrainHeight
osSetTerrainHeight
osGetSunParam
osSetSunParam
osSetPenColor
The functions that do not comply to the standard give a warning when used but work normally otherwise.
The graphics primitive drawing command "PenColor" has also been added as well as dynamic texture parameter "bgcolor" as an alternative to "bgcolour".
The following two functions have been renamed because they are not enabled yet aynway:
osWindParamSet => osSetWindParam
osWindParamGet => osGetWindParam
- fixes wild swings in memory usage related to usage of GetDrawStringSize()
We've been seeing wild swings in memory usage and a large chunk of
memory leak. From analysing this it's pretty clear that the mono
garbage collector is rather buggy! When exercised heavily it looks
like it frees more than its meant to resulting in crashes.
GetDrawStringSize() measures the size in pixels of text. To do this
memory for an image is allocated and used to call the GDI text
measure functions. Although no reference to the temporary memory
for the measuring is kept, it takes quite a while for the mono
garbage collector to clean up - so if lots calls to
GetDrawStringSize() are made at once there can be a spike in memory
usage. If the garbage collector is not fast enough then the GDI
layer runs out of memory. It also looks like the garbage collector
is not always reclaiming all of the memory.
I've attached an OpenSim patch which works around the garbage collector
issues. Instead of dynamically allocating memory for measuring
text sizes, it serialises (on a per region basis) access to a single
block of memory. The effect of this is to be nicer to the garbage
collector as it has a lot less work to do, at the cost of some
theoretical loss in performance (nothing noticeable with our tests
which hit it pretty hard).
OpenSim still does leak memory slowly, but it is a lot more stable
with this patch. I suspect that either the garbage collector misses
bits of freed memory or the GDI/cairo layer leaks a bit each time a
texture is created. Thats going to be a lot harder to hunt down, but
for reference if someone has OpenSim running on Windows it would be
interesting to see if it has the same problem as it would tell us if
its a mono/GDI problem or an OpenSim problem.
Adds a test to see if the first option on osDynamicTextureData is "AltDelim",
then picks up the first character after the whitespace and uses as a delimiter
instead of ;. If this string does not appear at the start of the data, the
default ; will be used, hence this should not break existing code.
