… reduce the dependency on the boost thread library

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…d inside ImageCacheImpl with a simpler idiom using a thread_local map

Signed-off-by: Chris Kulla <ckulla@gmail.com>

… are no longer used

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…This could lead to the same pointer being re-used and getting stale data from the previous cache's per thread info structs. Instead of looking up by pointer, lookup by an atomic id that is never re-used. Also remove the shared flag because it was always set to true. Either the thread owns the per thread info (in the case of automatic creation) or the caller down (manual creation). Both cases were being flagged as shared, so the non-shared codepath was never taken

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…e itself own (and free) the memory of all per thread info structs, regardless of if they are created implicitly or explicitly.

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

… destroyed.

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…rors are destroyed."

This reverts commit d7e1afe.

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…mageCache and TextureSystem alone as they are less likely to accumulate cruft and also to avoid the static destruction order fiasco

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…ve the ImageCache and TextureSystem alone as they are less likely to accumulate cruft and also to avoid the static destruction order fiasco"

This reverts commit ac5752c.

Signed-off-by: Chris Kulla <ckulla@gmail.com>

…can sort out the static destruction order fiasco

Signed-off-by: Chris Kulla <ckulla@gmail.com>