Re: tests code coverage for php 5.5

On Thu, Jun 5, 2014 at 4:44 PM, Maciej Sobaczewski <sobak@php.net> wrote: > On Thu, Jun 5, 2014 at 2:55 PM, Maciej Sobaczewski <sobak@php.net> wrote: >> >> On Wed, Jun 4, 2014 at 10:06 PM, Nicolai Scheer < >>> nicolai.scheer@gmail.com> >>> >>>> wrote: >>>> >>>> Hi, >>>> >>>>> >>>>> looking for tasks to invest a little spare time in the weeks to come I >>>>> had >>>>> a look at >>>>> >>>>> http://gcov.php.net/viewer.php?version=PHP_5_5 >>>>> >>>>> The test coverage is suspiciously low... >>>>> >>>>> Test coverage of 5.4 is quite high, so is coverage for 5.3. >>>>> >>>>> I would have guessed, that for a new major php version all tests from >>>>> the >>>>> previous version are reused, and fixed if needed. >>>>> >>>>> Looking at the 5.5 source tree all the phpt files seem in place... >>>>> Aren't >>>>> the tests just currently not executed on gcov.php.net? >>>>> >>>>> Thanks for any insight! >>>>> >>>>> Greetings >>>>> >>>>> Nico >>>>> >>>>> >>>>> Yeah, something is definitely wrong: >>>> http://gcov.php.net/PHP_5_4/lcov_html/ >>>> vs >>>> http://gcov.php.net/PHP_5_5/lcov_html/ >>>> >>>> and it we dates are correct, the last gcov build for 5.6 was more than a >>>> week ago: >>>> http://gcov.php.net/ >>>> http://gcov.php.net/PHP_5_6/lcov_html/ >>>> >>>> ps: I've cc'ed Nuno as he is the man who knows our gcov setup. >>>> >>>> >>>> I will ask by the way. Aren't any parts duplicate of what we've got >>> with >>> travis/QA reports now? I mean statistics about passed/failed tests etc. >>> It >>> is probably good to have all such data in one place, but not when tools >>> are >>> not working properly. It also means two systems to maintain and in my >>> opinion leads to confusion as the results are not equal. >>> >>> Regards, >>> Maciej. >>> >>> -- >>> PHP Internals - PHP Runtime Development Mailing List >>> To unsubscribe, visit: http://www.php.net/unsub.php >>> >>> >>> my idea with jenkins was to have everything in one place, so build and >> test >> php-src on multiple platforms(including windows) and also test userland >> projects with the resulting binaries and also have coverity reports (not >> for every commit, but periodically) but that was low prio, as gcov was >> already there: >> >> http://grokbase.com/p/php/php-internals/11btzq29x8/php-dev-ci-for-5-4 >> but stuff happen, and I never managed to finish what I was planning. >> having travis is a bit redundant, as it does less than what jenkins does >> (or can), and only supports linux with one distribution and only 64bit, >> but >> it doesn't really require any maintenance from our part and "just works" >> and also supports testing pull requests, which would require a bit more >> work on our part if we want to do that on our jenkins environment (not the >> PR building, but doing that in an isolated and secure way). >> > > Nice idea indeed. It would be good to have such tools in one place. > However, as far as I understand, it is abandoned now. > > > btw, I'm currently working on two pull request for our travis setup: one >> is >> for changing our current travis project from php to C, which saves some >> build time, as the travis box won't need to build php 5.4 which we doesn't >> use at all, and it will also mean that we will install our dependencies >> for >> ourselves explicitly instead of using those installed by default for the >> travis php 5.4. >> the other pull request is to do cross-compile 32bit builds on travis >> through ia32-libs. >> that would help us at least cover one platform/distro. >> >> > If we're currently focused on using travis ci maybe it would be good to at > least disable some duplicated functionalities from gcov.php.net? I'm not > certainly sure which have better and more up-to-date replacements now but I > thought about: > - compile results > - test failures > - expected test failures > - valgrind > - skipped tests - not sure it has replacement, but it seems that gcov also > supports only one distro in 64x architecture, so it's not a big help > > What do you think? > > Greetings, > Maciej. > I think it makes sense to see the number of executed/skipped/failed/etc. tests with the coverage report as they are directly related (if you add more tests that usually increase the coverage, if you remove some, the coverage will also drop). Plus we didn't really had any complaints about having that (most people doesn't even know about gcov). -- Ferenc Kovács @Tyr43l - http://tyrael.hu