Am 29.11.2013 02:26, schrieb 赵小强: > 于 11/12/2013 10:52 PM, Andreas Färber 写道: >> it would be nice if you could check whether these >> devices (the non-KVM versions at least) are covered by make check. For >> ICC bus I am certain that it is. > 1. Does "make check" mean the build target in the Makefile ?
Yes, in particular the make check-qtest target (check-qtest-x86_64). > if it > is, I can not find anything about "icc_bus" under "tests" directory. Or > you refer to something else? IIUC there is no strict need for an icc_bus-test.c because any test using -M pc (or default for i386 or x86_64) is implicitly using icc_bus and I don't recall any MMIO or similar implementation for ICC. It was just a hot-pluggable version of SysBus. My question was, does this implicit testing apply to the other devices touched here as well or not. > 2. What does "enable the *APIC devices and have the realize functions > actually exercised?" mean? It may mean for instance, add printfs (or assertions or breakpoints or trace points in core QOM code...) to the initialization code of those devices, run make followed by make check to see if all of them show up. If not, add apic-test.c, ioapic-test.c etc. using whatever -device or -cpu command lines necessary to trigger the printfs - and then obviously drop any debug printfs agains. But if I knew the answer already I wouldn't ask you to investigate. :) What I am not asking for is functional tests for registers or whatever - you did not write that code, you should just assure that your realize refactorings do not break device initialization accidentally. If you do find time to work on this and find that some non-KVM device is not covered: I was going to split off icc_bus into its own patch from 2/4. Regards, Andreas -- SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg
