The May online Rougol talk saw the welcome return of Gerph, expanding on his previous talk with a more detailed look at testing.
Gerph's day job is to provide a tool for testing and debugging. Here are my notes (all mistakes are mine) on a fascinating talk.
Not having tests is unsettling.
The RISC OS build service was an idea to improve testing on RISC OS. So far has seen little usage from other developers.
Testing tends to be ad hoc on RISC OS. Not much automated testing.
Testing on RISC OS is hard. Lack of tools has been one issue.
Automated tests give you confidence that your software keeps working.
Industry standard figures for bugs are 10 defects per 1,000 lines of code would be very good.
Manual tests gives you confidence but does not really check your code.
Not all bugs matter - some have more impact than others.
Key measure of whether you have enough testing is whether your users are having issues.
It is common to miss things in testing.
Testing is anything which makes sure your product does what it is supposed to. It can include peer reviews and other things as well as running the code.
Gerph covered different levels of tests - Unit, Integration, System, System integration, customer tests.
There were some nice examples for test code shown. These included a nice demonstration of ErrorCancel being tested in pyro and some module tests.
A nice feature of tests written in Perl (and now Python) was that the same code can run on Linux, MacOS and RISC OS.
Gerph still feels RISC OS too easy to be borough down by rogue code - one of the reasons for writing Pyro.
The demonstration of testable kernel components was very impressive!
If you are interested in testing, there was a lot of really interesting material above and beyond RISC OS on testing in general which you would find very useful. I will be rewatching the video when it comes out t wget some ideas for my day job work.