Bug#1131997: llama.cpp: Frequent crashes with Qwen3-coder

[Philipp Klaus Krause]

Am 31.05.26 um 15:45 schrieb Christian Kastner:
could you try reproducing this with the newer llama.cpp/ggml in testing?
I can not reproduce the bug using llama.cpp from current Debian testing; 
looks like this is fixed now.

Philipp