On Tue, 22 Dec 2009 22:48:35 -0500, Jonathan Yu wrote: > I know gregor prefers us to run as many tests as possible; however, in > light of their arguments, I now think it's safer for us (not to > mention less maintenance work) to just run AUTOMATED tests, and no > longer run RELEASE tests, except in rare occasions. If an upstream author clearly states "these tests are for me and those tests are 'public'" -- be it by separating t/ and xt/ or by using both AUTOMATED_TESTING and RELEASE_TESTING -- I'm happy to follow their intention. Difficulties arise when there's either no clear separation or when one environment variable needs to be set to run any test -- in these cases I think we're back to deciding on module-by-module basis (and taking our experiences into account, i.e. don't run Perl::Critic tests). > I propose we drop the "run RELEASE_TESTING tests by default" and > consider each of these modules carefully. At the moment I think we have to look at _which_ tests are de/activated by the RELEASE_TESTING flag, there doesn't seem to be common approach yet. But in general I think that standardization on both syntax (which variables are used) and semantics (which tests are covered by which variables) of the test suites on the toolchain side is a good development, and once it's sorted out I think we should follow that. Cheers, gregor -- .''`. http://info.comodo.priv.at/ -- GPG Key IDs: 0x00F3CFE4, 0x8649AA06 : :' : Debian GNU/Linux user, admin, & developer - http://www.debian.org/ `. `' Member of VIBE!AT, SPI Inc., fellow of FSFE | http://got.to/quote/ `- NP: Mercedes Sosa: Fragilidad
Attachment:
signature.asc
Description: Digital signature