[forgot to cc this thread] On Sat, Jun 13, 2026 at 05:34:16 +0000, RVP wrote:
> This is PR bin/59635, I think. I'm pretty sure this is fallout from the "offset" fixes. The observed randomness probably depends on whether `realloc` in `makemessage` has moved the object or not. I think I can force it to always crash by forcing the reallocation to get new memory. Someone who groks the internals should look at this, b/c I see omsgCount = 0, but omessage != NULL and I don't have time to learn what does that mean and how the conditions to call `thread_fix_old_links` should be tweaked. -uwe
