Re: Failing tests on Fedora

On Wed, May 28, 2014 at 8:00 AM, Pavel Raiskup <praiskup@redhat.com> wrote:
> On Tuesday 27 of May 2014 21:54:07 Heikki Linnakangas wrote:
>> On 05/23/2014 08:32 AM, Pavel Raiskup wrote:
>> > Attached files (regression.out, regression.diffs).
>> > *** /home/praiskup/rh/projects/psqlodbc/test/expected/deprecated.out        2014-04-02 08:37:57.065266936 +0200
>> > --- /home/praiskup/rh/projects/psqlodbc/test/results/deprecated.out 2014-05-22 19:01:21.702283366 +0200
>> > ***************
>> > *** 6,12 ****
>> >   Check for SQLGetConnectOption
>> >   Access type is: read-write
>> >   Check for SQLError
>> > ! Error check: ERROR: relation "table_not_here" does not exist;
>> >   Error while executing the query
>> >   Check for SQLSetParam
>> >   Result set:
>> > --- 6,12 ----
>> >   Check for SQLGetConnectOption
>> >   Access type is: read-write
>> >   Check for SQLError
>> > ! Error check: ERROR:  relation "table_not_here" does not exist at character 13;
>> >   Error while executing the query
>> >   Check for SQLSetParam
>> >   Result set:
>>
>> Hmm, you get that "at character" appendix in the error message if you
>> connect using the FE/BE protocol version 2. Is there some middleware or
>> something else in your test environment that you didn't mention?
>
> No, I had just improperly configured testsuite.  Thanks for pointer -- I
> should use 7.4 (v.3 protocol) rather than 6.4 (set in odbc.ini in git
> master).
>
> After that change, only two failures seem to occur:  First is relevant
> to mail thread discussing the commit 56b18326 && ad2a4aecb4.  The second
> one is relevant to 'prepare' test:
>
>     @@ -109,8 +109,8 @@
>      4098   4098
>      4099   4099
>      Result set:
>     -1      @ 1 day one day
>     -2      @ 10 secs       ten secs
>     +1      1 day   one day
>     +2      00:00:10        ten secs
>      # of result cols: 2
>      Result set:
>      2      bar
>
> Even when I see this was here before, it is pretty unlucky we are unable
> to run testsuite against older PostgreSQL servers than 9.4;  it is common
> practice (I would not say that if the current git branch was supposed to
> be run strictly against 9.4).
>
> To make the testing little bit easier, test/odbc.ini could enforce V3
> protocol by default.  But it is OK for me from now :)
Yeah, I'd agree with that, especially knowing that the plan is still
to clean up code with protocols older than 7.4 for the 9.4 series that
should show up in a couple of months. Now still seems too early.
--
Michael