While testing #7687 we discovered, that the nucleo-l152 (stm32l1) is broken since quite a while now. Bisecting lead to the changes in #7385 (specifically to dd49f22).

The change from __irq_enable() to irq_restore(state) breaks the L1 for some reason, can't really tell why...
When doing s/irq_restore(state);/__enable_irq();/, everything is working (again). The difference seems to be /__ASM volatile ("MSR primask, %0" : : "r" (priMask) : "memory");
vs
__ASM volatile ("cpsie i" : : : "memory");

Any ideas?