On 20.03.2018 14:17, Christian Borntraeger wrote: > David, Jason, Michael, > > the cpumodel code is somewhat fragile as we have to add maintain things > in multiple places. I would like to have more robust code, e.g. by either > generating more or by having build bug_ons or something like that. > Any idea is highly welcome. > > Christian
Yes, error prone. I was also wondering if we should explicitly address array items when initializing - e.g. in target/s390x/cpu_features.c the "indexed by feature number for easy lookup" part. More LOC but less error prone. As Mimu also suggested, if we can generate something automatically, we should do that. -- Thanks, David / dhildenb
