Re: [PEAR-DEV] Re: BC breaks: Cache_Lite, PHP_Documentor

On Fri, Sep 10, 2010 at 10:32 AM, Daniel O'Connor
<[email protected]> wrote:
>> How do you like this:
>> http://friendpaste.com/1jrxQBie31IiJKXZDXaR0P
>>
>> (Not a huge fan of $GLOBALS, but it would keep the API.)
>>
>>
> It'll certainly get around the appearing-in-stats issue; and would probably
> work for a few others in the list. The unit-test-loving-me sighs because
> it's not a fix to the problem people are trying to deprecate; the
> don't-break-bc-me grins because it's dancing the fine line.

I don't understand half of what you're saying.

The suggested way is indeed $_SESSION['foo'] = $foo; instead of
session_register('foo'); ;-)

> It does seem pesky to make a HTTP_Session3 just to swap to $_SESSION.

Why? HTTP_Session2 is alpha/beta grade. I *could* break BC all I want
[that's my point] and the above patch doesn't even break it. ;-)

I see if I can add a couple tests to see how this behaves and then
I'll go ahead and squash it.

> If we're re-defining the problem as don't-appear-in-stats, Olle, how do you
> feel about making a "except these errors" filter? (yay, rug sweeping)
>

I kinda don't want to hide "errors".

But let's say a package is superceeded by another more modern
implementation, then (IMHO) there is no reason to pester the
developers or anyone else about E_DEPRECATED in that package. You'd
basically move on.

Till