On 2/22/10 1:33 PM, Peter Stuge wrote:
Stefan Reinauer wrote:
Also, do we want to die() on an assert?
Ideally I would like coreboot to be so structured that an assert just leads to the current "task" being aborted, and then we try to continue as best as possible.
Yes. I agree. Intuitively I'd think that this is better written down in code at the place where it happens, though, instead of trying to do magic in order to detect which scope to drop out to. If no magic is required, I'm all for doing this. In Forth or C++ or Java we could do a try { ... } catch (...) kind of thing.
It would be great to also have a panic room, but I agree that it should only be called fairly infrequently. Ie. I'd like to do away with die() in the long term.
I still try to fancy what that panic room is. Is it the gdb handler for the case that gdb is enabled?
Stefan