On 08/19/2017 09:34 PM, Philipp Kern wrote: >> So it's not kernel issue or anything else. It must be something wrong >> with installer itself. Should I report a problem somewhere? or we are >> only ones who can't do fresh install on hercules? > > I'm happy to retest if you can give me the hercules.cnf you used.
Bercik sent me their config and I'm able to confirm the bad behavior. But it looks to me as if console messages are swallowed. As soon as I add sleep() calls around all my writes, they succeed. (Which makes printf debugging very weird.) And indeed, if I enter ".1" it continues with "The following device numbers might belong to CTC or ESCON connections.". So all that happens is that either Hercules or the kernel while sending swallows the output and you can't actually see the prompts. That also matches the fact that there are very few startup messages from the kernel altogether. Essentially every message that doesn't have a bunch of other messages following (plus the first of the batch) lands on the console, the other ones are discarded. That makes me incredibly suspicious that this is a problem in Hercules. It's also not in the typescript generated by script(1), so it's not simply overwriting screen content either. I guess we'd need to either raise this with the Hercules people or instrument Hercules to see where the lines are missed. Kind regards Philipp Kern
Description: OpenPGP digital signature