Hi Miao, On Mon, Sep 21, 2020 at 12:21 PM Miao-chen Chou <mcchou@xxxxxxxxxxxx> wrote: > > Hi Luiz, > > On Fri, Sep 18, 2020 at 3:12 PM Luiz Augusto von Dentz > <luiz.dentz@xxxxxxxxx> wrote: > > > > Hi Miao, > > > > On Fri, Sep 18, 2020 at 12:44 PM Miao-chen Chou <mcchou@xxxxxxxxxxxx> wrote: > > > > > > Hi Luiz and Marcel, > > > > > > Unlike the rest of the existing unit tests in BlueZ, the logic blocks > > > tested in test-adv-monitor require dependencies of not only > > > src/adv_monitor.c but also all the dependency tree of > > > src/adv_monitor.c. The current convention in Makefile.am is to add all > > > the extra dependencies one by one. However, the maintenance cost is > > > high and not suitable in the case of test-adv-monitor. Therefore, we'd > > > like to propose changes in Makefile.am to make the source of > > > bluetoothd as a static library and link it for bluetoothd target and > > > the unit test target. It would be great if you can provide feedback on > > > this idea before the implementation. Thanks in advance! > > > > Then we should have had the code move to src/shared for unit testing, > > but how exactly are you planning to do that? For testing the kernel > > interface it normally done via a dedicated tester, but that again can > > be done with shared library. > > > In series https://patchwork.kernel.org/project/bluetooth/list/?series=351021, > we introduced some helper functions in adv_monitor.h to perform unit > testing and test-adv-monitor to facilitate the unit tests of > adv_monitor. We are encountering an expected build check failure on > this series. > > There are two categories in test-adv-monitor, content filtering and > RSSI tracking, and content filter is easy to be moved to a standalone > shared component while RSSI tracking involves the use of timer, D-Bus > method calls and adv_monitor's internal structures, and that makes it > strongly coupled with the adv_monitor implementation which require a > tree of dependencies apart from adv_monitor. > > There are two options to resolve the build failures in our case. > (1) Reorganize Makefile.am > This option is to make the sources (except src/main.c) into a static > lib and link this lib in bluetoothd executable target and whichever > unit test the sources are required. > (2) Create src/shared/am to facilitate helpers and core logic > This option is to create a new source under src/shared/ to facilitate > helper functions and core logic for src/adv_monitor. The interface of > src/shared/am may have the following functions. > - Create/destroy functions of struct adv_monitor > - Create/destroy functions of struct adv_monitor_device > - Helper function for monitor content matching. > - Helper function for RSSI tracking > However, the logic of RSSI tracking is hard to be ripped off and moved > to src/shared/am. One example would be the use of timer in RSSI > tracking, and there is currently no previous example of the timer use > in the shared library. Not really following regarding the dependency on D-Bus, usually shared don't have dependency on that because the code would be part of the deamon and in that case you would be better of testing that direct via a tester that does exercise its D-Bus API. Lets be clear here, except for gbus itself all our tests under unit directory are for C function testing, if you want to test a D-Bus interface then we need a tester that would do that over D-Bus. > Series https://patchwork.kernel.org/project/bluetooth/list/?series=351021 > is up for review. Our next step here would be ripped off the unit test > for now and submit v5 of the series. Once we reach an conclusion on > advmon unit test, we can submit a separate series including the > refactoring and unit tests. Looking forward to any feedbacks. Thanks! I'd strip the testing for now since it doesn't seem we are on sync to how we test things apparently. > Regards, > Miao -- Luiz Augusto von Dentz
