On 18/07/2018 20:29, Emanuele wrote:
I had to put this patch here because it also introduces
qpci_device_init, used by sdhci (patch 3).
For the next version I plan to have a patch X where I rename all
occurrences of qpci_init_pc in qpci_pc_new, and a patch X+1 that
introduces qpci_init_pc (was qpci_set_pc) and the other changes.
Should I only introduce qpci_device_init in patch 3 and the remaining
things in patch 5?
I think the general problem here is that in some patches I create
functions that are planned to only be used only in next patches (of the
current series).
I think it's okay this way, however you should justify the changes you
make to "qgraph-ify" each component.
For patch 1, let's wait for Stefan's reply. Because patch 1 is
introducing the infrastructure, I think it is acceptable that some
definitions are introduced early as long as they have doc comments; it
would make little sense to introduce get_device in patch 4 just because
there are no "contains" edges until then.
However, introducing the qos-test directly at the beginning is also a
possibility.
In either case, we need better doc comments for the function pointers in
QOSGraphObject.