Re: Comments on bugs marked bogus

On Wed, 2005-07-27 at 09:13 +0200, Bertrand Mansion wrote: > Alan Knowles wrote: > > >I think autoload will be fixed to allow failure - It's a known bug of > >the autoload implementation .... > > The bug is not that it should allow failure. The bug is more that is_a() > shouldn't IMO call __autoload(). quite agree, using __autoload is like driving without headlights at night.. If you are lucky it might work.. - but as long as I'm in a different country, you are free to do it ;) > It's not as trivial as just changing one character in HTML_QuickForm. is_a() is > used all over the place in other packages too. I think a pragmatic approach here is best.. PEAR doesnt recommend __autoload(), use at your own risk, however if you find a 'is_a()' case issue, we will make that change.. > > Alexey has one point with PEAR_Error. This should be fixed first. We also have > an HTML_QuickForm_Error object, and like with PEAR, it is bundled in the same > file as HTML_QuickForm. >From what I remember, there was always 'edge cases' where putting multiple classes in a single file is more sensible.. (Exception class'es come to mind). So that will bugger up autoload whatever you do... > > __autoload() will force us to use one file per class. IMO, this is silly, > especially knowing that the biggest performance issue with PHP AFAIK is related > to seeking files. I think these issues are being addressed (or pretty much have been) addressed by core anyway. And to be honest there are usually an infinate number of other optimizations before you start having to deal with this.. (and if it was a real issue, there are better ways than autoload to deal with it - include callbacks/include call caching..) > > But I don't want to fight windmills. So the question really is whether we want > to allow the use of __autoload() with PHP knowing that, like magic_quotes, it is > not a good practice, or if we simply don't care. This decision will have > consequences on at least 50% of our packages, especially the big ones. > > The issue in HTML_QuickForm will be fixed when the RFC will be proposed and > approved by all. I guess if you are not willing to go with a pragmatic 'I'll fix a change request for case senistive is_a()'s , then someone who wants this change doing an RFC is really the only option, but it does seem alot of fuss over a tiny issue... Regards Alan > > Bertrand Mansion > Mamasam >