Attila Szegedi wrote: > > On 2008.01.16., at 9:54, Marc Guillemot wrote: > >> Hi, >> >> can someone answer following questions concerning Rhino sources. This >> would help me in providing patches: >> >> - has Rhino a valid test suite? >> Running the tests as documented at >> http://developer.mozilla.org/en/docs/Running_the_Rhino_tests leads to >> some errors and can't therefore be used to know if changes introduce >> errors. > > Which ones are failing for you? I'm running my tests on Mac OS X, with a > non-english locale, and I too have some tests failing. Namely: > > ecma_3/Date/15.9.5.5.js: says it'll fail for a non-english locale, and > my machine indeed runs with Hungarian, so that's probably okay > > js1_5/Regress/regress-58116.js: this might be genuine, I'm not sure > > js1_5/Regress/regress-98901.js: throws StackOverflowError. From reading > the associated bug, this is probably (unfortunately) expected. They > fixed it for SpiderMonkey, but it is caused by a hideously ugly piece of > script... I guess we might just want to add it to our exclude list and > forget about it :-) > > lc2/JavaToJS/char-002.js: it is simply not platform independent. I don't > even think this test is correct. It presumes that the Java path > separator character is ; (semicolon), but on UNIX-like OSes it is > actually : (colon). Therefore, this test fails on all non-Windows OSes. > > I discussed this with Norris back in December, they work for him.
grep Failed on the output gives me following failures (meaning that stackoverflow problems are not included here). It differs a bit from yours: ------------------------------- Failed: ../tests/ecma_3/Date/15.9.5.5.js: Math.abs(Date.parse(now.toLocaleString()) - now.valueOf()) < 1000 = false FAILED! expected: true Failed: ../tests/js1_5/Regress/regress-58116.js: FAILED! expected: Expected value '-120', Actual value '-60' Failed: ../tests/js1_5/Regress/regress-89443.js: JavaScript errors: Failed: ../tests/js1_5/Regress/regress-98901.js: JavaScript errors: Failed: ../tests/js1_5/extensions/regress-226507.js: JavaScript errors: ------------------------------- In fact I don't really care if these tests are excluded or not but it would help if the tests that should pass could be up to date (additionally a mechanism to detect tests that already pass even if they are marked as excluded would be useful). >> - is there a continuous integration for Rhino somewhere? > > I started setting one up at <http://cn-rhino01.nl.mozilla.org:8085>. > Problem is, there are still some quirks with it. One of them is we're > using the Mozilla anonymous CVS (as I don't want to store anyone's > developer credentials in the system), and it can't detect commits > automatically for some reason. Also, it must check out the mozilla/js > tree, not the mozilla/js/rhino tree, as it needs mozilla/tests in order > to run. > > The biggest problem is that the JUnit test driver produces even more > failures than the command-line driver, so we're yet to produce a > successful build (it reports 507 failed tests out of 4647) :-( do you know where the difference comes from? Working on java.util.regex based regex support for Rhino I've seen that some problems occur due when tests are run in a separated thread which has a smallest stack size. Perhaps is it the same problem here? >> - are there some benchmarks validating the choice of some >> "optimizations"? >> I'd like for instance to try to use standard java collections rather >> than self made hash table to store Slots. > > That's admittedly one of my pet peeves too. Rhino codebase goes back to > Java 1.1 days when the only collection was java.util.Hashtable. It'd be > due to be modernized, certainly. does it mean that there is no benchmark and that passing the tests would be considered as ok? Cheers, Marc. -- Blog: http://mguillem.wordpress.com _______________________________________________ dev-tech-js-engine-rhino mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-js-engine-rhino
