RE: [PEAR-QA] BC check on release upload

From: Date: Wed, 17 Sep 2003 09:56:54 +0000
Subject: RE: [PEAR-QA] BC check on release upload
References: 1  Groups: php.pear.qa 
Request: Send a blank email to pear-qa+get-194@lists.php.net to get a copy of this message
> From: Wolfram Kriesing [mailto:wolfram@visionp.de] > Sent: Wednesday, September 17, 2003 11:53 AM > Lukas Smith wrote: > >>From: Pierre-Alain Joye [mailto:paj@pearfr.org] > >>Sent: Wednesday, September 17, 2003 9:34 AM > > > > > > > >>While discussion about Log 1.7.0 which introduces a BC break between 2 > >>releases with the same major version number. I started to work on a > >>interface checker. > >> > >>It checks the functions/classes/class::method between the uploaded > >>release and the previous release, only if the major version number is > >>the same. > > This is very good stuff. > > However what about private methods, classes etc? > > > > Therefore I would not rejected the release but issue a warning > > containing the differences that were detected. The maintainer then needs > > to explicitly say: "no I still want to upload because I know what I am > > doing". > > i would agree with Pierre, that the package should not be uploadable. > Some other packages might depend on it, the dev might know what he is > doing, but if other people can not rely on getting the same API for > sure, i would not allow upload. The dev should mark methods as > deprecated, within one major release, he can remove the method(s) for > the next major release. > > Aehm, i think version <1 (0.x) should not be checked, right? Since > they cant be stable anyway ... if i understood it right. I know there > are packages that claim to be stable and are in 0.x (i got at least > one too :-( ). The thing is that a class might be privat. Until we also check the phpdoc comments we will never know that. Furthermore some functions might have been tagged as private with phpdoc comments but not prefixed with an underscore. Anyways making this such a hard rule removes common sense. If we don't trust the developer to know what he is doing when he presses "yes I am aware of these changes and they don't result in a BC break to any public class/method/function" then we might as well forget about PEAR development model itself. Regards, Lukas

« previous php.pear.qa (#194) next »