>> Much-to-most of this "B2G-only" code compiles on other platforms, too. >> Or at least, that is true of the B2G patches I've reviewed. We >> shouldn't exaggerate the effect of this without some data. (Note that >> basically any code we want to write unit tests for at the moment must >> compile on more than B2G, so it's not entirely unintentional that so >> many patches touch other platforms.) > > If that's the case, would pick-a-single-platform (whether -p any, or > done manually) be sufficient for testing? > > If it is (that is, if the developers are comfortable not having a > B2G try build at all, but want the try build only to check the unit > tests), then that's likely what should be happening, unless these > unit tests have platform-specific variations between the desktop > platforms. > > If it's not (because the developers want to test one platform where > the unit tests run and test that B2G builds), then that bug is still > key to reducing load.
You're right: -p any,b2g is probably sufficient for many patches, but we really should test that B2G builds, since that's the only CI test we have for B2G on tinderbox at the moment, and since we do have a fair bit of b2g-only code. But that means we need -p b2g. _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
