Do no worry nobody is offended. :-) We can do better and we should make better. Anyway let me summarize the status:
- There was a problem building OSX and Windows binary. This was because of the different folder structure in those distributions. This problem should have been caught long ago. I am surprised we did not. Here we can do better. (*) - Remember we changed 57000 lines of code. As a result we had a few backward compatibility issues, we test a lot but we cannot test everything (default adapter in firebird, unable to create self references when the id field is renamed, markmin failure on ````` strings). These are minor issues that are expected. Affect a small minority of users. So far we have addresses them all within 24 hrs. - We had some false positives. Bug reports that seem bad but have nothing to do with 2.0.x and therefore confused the issue. They are pre-existing conditions or bugs in user code. - There is also a problem with upgrades. The automatic upgrade from admin does not detect the presence of the new 2.0.x version. You have to upgrade manually. This was a bug in 1.99.7 not in 2.0.x. We are thinking about how to address it but we do not have a solution. We may be able to fix it by calling 2.0.3 as 1.99.8. Yet does not seem like a good idea. The real issue was (*). Because of that I have rebuilt the windows and mac distributions with slightly different versions than the source distribution using a later web2py version. If the problem is resolved, tomorrow, I will repost 2.0.4 and make sure everything is again in sync. Massimo On Friday, 31 August 2012 13:08:41 UTC-5, viniciusban wrote: > > +1 > > Again, note that I'm not against our team. I'm on it, too. > > I'm not criticizing. I'm suggesting quality improvement. > > -- > Vinicius Assef > > > > On 08/31/2012 11:07 AM, Richard Vézina wrote: > > Hello, > > > > I love web2py and normally I just wait a bit when there is new version > > to make sure I don't spend to much time on version issue. > > > > I also test de trunk once in a wild, but I found difficult to proceed > > that way. 1) I don't have much time, 2) there is often little issue that > > will be gone in a release that are actually artifact of development I > > think. I mean, the developer know that there is most probably something > > wrong with the new code he produce, but just can't test it in all > situation. > > > > For sure, tag a beta testing version before releasing could make more > > work and struggle with version control to make sure patch spread over > > all branch (trunk, beta), but I think it could be very good to have a > > beta testing a week before stable. > > > > Anyway, I don't have to much problem with the actual practices, but I > > think beta test could had avoid 1.99.5 for example, since this version > > has be the beta test version and 1.99.7 the final. > > > > Richard > > > > On Fri, Aug 31, 2012 at 9:49 AM, Anthony <[email protected]<javascript:> > > <mailto:[email protected] <javascript:>>> wrote: > > > > > Web2py has so much features and it is impossible to test > everything. > > > > Sorry, but I cannot agree with this statement. Is it the > > official position? > > > > > > Well, we can probably improve the unit test coverage. And there's > > been recent talk of developing a test application to check against > > for some functional testing. But as with any software (particularly > > as complex as a web framework), it's not possible to test every > > conceivable permutation of functionality a user might implement. We > > have to rely on reports of bugs found in the wild to some extent. > > > > "Stable" means: "you can download and just use it". Anything > > different > > is "almost stable" or "buggy yet" or, using beautiful words, > > "release > > candidate" or "pre-release version". > > > > > > We had two (officially labeled) release candidates > > ( > http://code.google.com/p/web2py/source/detail?r=ed41a29eb7c2e283587c141d0464b6c9be68eb0d). > > > > Maybe we should change the "Nightly Build" label on the downloads > > page to "Release Candidate", and perhaps advertise a bit more. Not > > sure it will help, though, as there were already many requests for > > testers. What do you suggest? > > > > Anthony > > > > -- > > > > > > > > > > -- > > > > > > > --
