3 ms·
One thing that wonders me. The patch contains a fix but obviously no unit test that verifies it. Is that the standard of linux kernel development? In the compan
by micw 3y ago
One thing that wonders me. The patch contains a fix but obviously no unit test that verifies it. Is that the standard of linux kernel development? In the company I work for, the patch would certainly be rejected because of that.
- romland 3y agoNot too long ago, this was actually discussed on HN here: https://news.ycombinator.com/item?id=33742130 https://news.ycombinator.com/item?id=33742130
- toyg 3y agoTbf the maintainer said the behaviour showed up in one of the existing tests, so effectively they already had it.
- pjmlp 3y agoUnits tests just like extreme programming, it was something that came and went, and only a few companies keep on doing. On most of project deliveries we have gone back to only integration testing.
- hnfong 3y agoHow do you unit test a #ifdef CONFIG_PPC32 ? I'm unfamiliar with how the Linux kernel does their tests, but presumably it would require a diverse set of hardware, and a lot of code would have to be in the form of integration tests...
- yencabulator 3y agoA very large portion of the Linux kernel code can't even run in userspace. It's not your average software project. Here's one integration test project: https://linux-test-project.github.io/ https://linux-test-project.github.io/ -- typically (non-platform) subsystem maintainers write their own stress tests. But stress tests are merely probabilistic.