Often new ebuilds fail during the test phase, but have hardly problems otherwise. My current strategy is to use maketest end retry failing ebuilds without maketest. Perhaps this could be automated avoiding a recompile: Try emerging with maketest enabled if it fails continue anyway, but display a warning at the end with a list of ebuilds with failed test phases. Reproducible: Sometimes Steps to Reproduce: 1. FEATURES="maketest" emerge =kde-base/kdelibs-3.3.2-r8 for example Actual Results: test phase failed Expected Results: tried testing, but continued and warn about the failing test phase
What's the point of running the test at all if the package is going to be installed to the live system? I don't really see the point in this. If you'd like this option so that emerge will continue installing other packages, the plan for the future is for emerge to continue with other branches of the dep graph. In other words, emerge will continue until all remaining packages depend (directly or indirectly) on any failed package(s).
Just trying the tests, you can find out problem quicker and know installed packaged that might not work smoothly. Just skipping over failed tests sound reasonable too.
know how ? the tests will run and fail and in a typical emerge, the user will never notice because they dont watch the output ...
right, but as of now, make test is mostly broken. i just did `FEATURES="maktest" emerge -e world`, and it failed on many packages that otherwise are fine (major packages such as cups too). so... it might be a good idea for testing purposes
If test's are failing... hows about filing bugs for the packages that are failing? This request seems a bit like noting that your tire has a nail in it, but wanting to ignore it...
i'm not sure i agree... i recently did `FEATURES="test" emerge -e world` and about 1/50 of 500 packages failed test. afaict, all of those packages still work. some of them i use every single day, yet make test said they were broken... the feature is still far too experimental on gentoo to be accuratly used to test imho. most maintainers don't even bother executing make check before they commit something, even if they are calling it "stable"
i'm with Brian on this if a package fails test, then fix the package