Cool VL Viewer forum

Hi Henri,

got another crash.

Windbg says something about: "ExceptionCode: c0000094 (Integer divide-by-zero)"
and maybe heap corruption.

Logs & crash dump attached.

I was unpacking some item from a worn attachment and right mouse clicked it. Not sure if this has anything to do with it.

Sadly, this is v1.32.2.11, and I do not have the symbols file for it any more, so that crash dump cannot be exploited (I wish Windows could produce plain text stack traces too, like Linux and macOS do).

However, if this was a crash after a right-click on a rezzed item (attached or not), then you likely simply hit the known selection outline crash bug in v1.32.2.11, that has already been fixed in v1.32.2.12 (see the change log).

Please update to v1.32.2.12...

Thanks Henri.

Guess i looked too early for the 1.32.2.12 version (around 12:00 CEST or so on 7.9.), so missed the update. But nice the crash is already gone again.

Well, in fact I noticed that crash earlier in the week with clang builds, but could not reproduce it with gcc builds, and the stack traces were all over the place, making it extremely hard to diagnose (in fact, none of the stack traces I got was pointing to the actual, culprit code), and that's only when I tested the v1.32.2.12 Linux aarch64 build on my ARM64 SBC that I could reliably reproduce the crash and conclude it was due to the new selection outlines code, at which point I reverted that code change in v1.32.2.12 and rebuilt everything...

This of course induced a delay compared with my "normal" release schedule: such things happen and you should not rely on a release every Saturday at 12:00:00.00 CE(S)T !

On a side note, I since found the bug in the new selection code which is now working fine: it was due to an improper (too small) std::vector sizing for some primitives and, depending on the compiler and how it reserves room for vectors (i.e. how much extra space it allocates on each resize() or reserve() calls), the bug resulted in an out of bound access (and memory corruption) or not... So, that new (fixed) code is part of next release already.