On Wed, Jan 21, 2026 at 08:21:42AM +0300, Dmitry Antipov wrote: > On Tue, 2026-01-20 at 14:55 -0800, Darrick J. Wong wrote: > > > Yes. Common code needs to have a rigorous self test suite, because I > > see no point in replacing inadequately tested bespoke parsing code with > > inadequately tested common parsing code. > > Nothing to disagree but: > > 1) My experience clearly shows that it takes a few patch submission > iterations and a bunch of e-mails just to notice that the tests are > mandatory for lib/ stuff. If it is really a requirement, it is worth > to be mentioned somewhere under Documentation/process at least.
Feel free to submit an update! :-) Sorry that I mentioned it one or two versions later than I should have. > 2) I've traced memparse() back to 2006 at least, and (if I didn't miss > something) there is no actual tests for it since them. And it's hard to > see a point in testing memvalue() prior to testing its actual workhorse. Yes, the historical code needs test cases. I added a few for get_option*() for example before touching that code. So you're welcome to start test cases for memparse(), I will appreciate that! -- With Best Regards, Andy Shevchenko
