[Bf-committers] "off by one" error in renderdatabase.c?

[Martin Dickopp]

Ton Roosendaal <ton at blender.org> writes:

> I tried rendering (OSX 10.3), no errors.
>
> The Blender code you suspect also has no evidence of 32/64 bits issues,
> and is in use 1.5 year in Blender without giving bug reports (if you
> were right we would also have issues in 32 bits Blender versions).

Hi,

I now have a sample file which, while not causing a segmentation fault,
generates the following messages on my 32 bit GNU/Linux system when the
scene is rendered:

  Memoryblock free: attempt to free illegal pointer
  Memoryblock Blove: end corrupt
  Memoryblock No name found: is also corrupt

This happens with the official 2.36 binary release from the Blender
website.

The sample file is here:

    http://www.zero-based.org/blender-bug-report/262144.blend

(warning: quite large, almost 11 MB). It simply contains a mesh with
between 1023 * 256 and 1024 * 256 vertices and therefore triggers the
behaviour I have described in my other posts. Instead of downloading
11 MB, the problem can probably also be reproduced by creating a mesh
with between 261888 and 262144 vertices and rendering it.

Can anyone confirm the behaviour I see? If so, should bug #2276 be
reopened?

Thanks again,
Martin