Lester Caine ha scritto:

> Umberto Masotti wrote:
>
> I've finally caught up with you!
>
>
>> Two test fails: 003.phpt and 006.phpt.
>> 003 is a false negative, because Fb2.1 has changed default field naming
>> in a more consistent way. They are no more called "FIELD_nn" but
>> "CONSTANT", "COSTANT_01", ... "COSTANT_nn" or "SUM", "SUM_01" ecc. ecc.
>>
>
> I'm happy to change the test, except that FB1.5 will produce the 'FIELD'
> results so perhaps we need a check for FB version and cover both cases.
> I've modified my copy and the test is clear ...
>

Maybe clearer duplicate the test, in front of each test search for
server version:
|ibase_server_info($service, IBASE_SVC_SERVER_VERSION)
SKIP (possibly with info output) the test if not server version expected.

|

>> In 006 last query return one more row in last "select... from.. between
>> ? and ?" where table data are integers between 0 and 9.
>> Parameters tested are "5.0" and 7.5. The resulting query then is similar to
>> select id from table1 where id between "5.0" and 7.5
>> Result expected is three ids: 5, 6 and 7. Result obtained is four: 5,
>> 6, 7 and 8.
>> Suspecting that 7.5 is converted as integer when assigned to parameter
>> (or something equivalent) so 7.5 becomes 8. I remember something has
>> changed in recent past about float to integer conversion but not for sure.
>> I feel this is another false negative.
>> Can anybody confirm?
>>
>
> I think this is the result of some other changes in PHP. There has been
> a discussion about 'rounding' being different, but I don't know if this
> is the same problem. '7.49' in place of 7.5 cleans the test up, but it
> depends on what we are actually testing for, and I feel that if 5.2
> produces a different answer to 5.3, then PHP is wrong?
>

Searching tracker and devel list I've not found anything relevant. Maybe
effectively some PHP problem.
Don't remember if same tests in 5.2 some time ago went OK, but I can try
with D4PHP environ.

Ciao.
Mimmo.