hi,
On Sun, Sep 4, 2011 at 7:52 PM, Stas Malyshev <[email protected]> wrote:
> So you're saying breaking BC is OK if only you can find some unaffected
> applications, right?
No, just like what I said for the is_a change, but it was acceptable
for is_a, right? is_a actually broken many (many) apps and codes out
there and it was easily identified, and it was even acceptable in a
patch release. Go figure.
>> There are incimoatibilities too across libmysql versions and across
>> mysql servers, which are actually affecting existing codes.
>
> I don't know of any incompatibilities that change semantics on this level,
> and anyway libmysql is beyond our control, but mysqlnd isn't.
Look at the history for the mysql's bugs, plenty of them, internally
or in userland.
>> I cannot find any code out there relying on this test case and I very
>> much doubt there is any. It is a non issue (unlike the is_a change for
>> example).
>
> Why we're testing for functions that, as you claim, nobody ever uses?
Can you please, and seriously, that's my last attempt, read my answers
and see what we actually test? I never claimed that nobody uses this
behavior but I cannot find any code, app or framework failing with
mysqlnd. Please find one and I will change my mind.
Btw, I have seen many totally pointless tests written for the sake of
having a better code coverage while the code itself makes absolutely
no sense or could even represent something that should not work this
way in the 1st place.
Cheers,
--
Pierre
@pierrejoye | http://blog.thepimp.net | http://www.libgd.org