Re: [PEAR-DEV] BC breaks: Cache_Lite, PHP_Documentor
[email protected] (Chuck Burgess)
| Newsgroups | php.pear.dev,php.pear.qa |
|---|---|
| Message-ID | <[email protected]> |
Given that phpDocumentor is an app rather than a library, does this change actually pose a risk of "breaking a user's code"? I'm inclined to think this is a safe enough change, albeit one that might highlight to the user that a php.ini setting should be reevaluated. Or, would a preferable alternative to the symptom (deprecated function) be to just switch to using ini_set()? CRB On Tue, Aug 24, 2010 at 12:37 AM, Matthew Fonda <[email protected]>wrote: > On Mon, Aug 23, 2010 at 8:14 PM, Daniel O'Connor > <[email protected]> wrote: > > To fix the following: > > http://pear.php.net/bugs/bug.php?id=17774 > > http://pear.php.net/bugs/bug.php?id=17690 > > > > requires a BC break for PHP4 users with magic quotes on. > > > > I propose: > > 1. Tag/Branch the code at this point, fix the issue on trunk for both > > packages > > 2. Put in the world's loudest "we just did a BC break" changelog > messages > > in the relevant releases > > 3. Publish that on the package homepage too. > > This is explicitly prohibited by PEAR standards [1]. A user may miss > the BC break notice, and find their code unexpectedly broken after > upgrading a stable package. People should not be using PHP4 still, but > unfortunately some are and we can't break things for them. > > I think this gets back to the unresolved discussion previously brought > up about old packages [2]. I like the "middle ground" idea Daniel > proposes, but as Helgi notes changing package channels is currently > not supported, so maybe we can find some other solution that still > allows us to move away from PHP4. Some PEAR usage stats on PHP > versions would definitely be useful to see. > > [1] http://pear.php.net/manual/en/rfc.version-naming.standard.php > [2] http://marc.info/?l=pear-dev&m=127280685832232&w=2 > > Best Regards, > --Matthew > -- CRB