Horton
Menu

Thread

  1. 7.3.2 Regression Failures Worth Looking Into

    Thomas T. Thai <tom@minnesota.com> — 2003-03-29T21:32:09Z

    NetBSD 1.6 / Alpha (64-bit)
Postgresql 7.3.2 & 7.4-dev
gcc-3.2.2

I compiled 7.4-dev and had success with all regression tests. However when
I went back to 7.3.2, about 15 out of 89 tests failed.

I'll be using 7.4-dev, but I'm wondering if it's worth it to the
Postgresql community if I dig further to see why those tests failed;
considering that obviously 7.4-dev fixed all of them.

--
Thomas T. Thai

  2. Re: 7.3.2 Regression Failures Worth Looking Into

    Tom Lane <tgl@sss.pgh.pa.us> — 2003-03-29T23:29:06Z

    "Thomas T. Thai" <tom@minnesota.com> writes:
> NetBSD 1.6 / Alpha (64-bit)
> Postgresql 7.3.2 & 7.4-dev
> gcc-3.2.2

> I compiled 7.4-dev and had success with all regression tests. However when
> I went back to 7.3.2, about 15 out of 89 tests failed.

Failed how?  Without looking at the diffs, it's impossible to evaluate
the seriousness of the issue.

> I'll be using 7.4-dev, but I'm wondering if it's worth it to the
> Postgresql community if I dig further to see why those tests failed;
> considering that obviously 7.4-dev fixed all of them.

AFAIK there are not any portability enhancements in 7.4 that might
explain such a difference. So the 7.3 results are interesting.  You
should probably rule out pilot error though (does "select version()"
confirm that you rebuilt with the newer compiler?)

			regards, tom lane

  3. Re: 7.3.2 Regression Failures Worth Looking Into

    Thomas T. Thai <tom@minnesota.com> — 2003-03-30T04:20:47Z

    > "Thomas T. Thai" <tom@minnesota.com> writes:
>> NetBSD 1.6 / Alpha (64-bit)
>> Postgresql 7.3.2 & 7.4-dev
>> gcc-3.2.2
>
>> I compiled 7.4-dev and had success with all regression tests. However
>> when I went back to 7.3.2, about 15 out of 89 tests failed.
>
> Failed how?  Without looking at the diffs, it's impossible to evaluate
> the seriousness of the issue.

I can send the diffs to you Tom, but it's too big to send to the list.
Would that be ok with you?

>> I'll be using 7.4-dev, but I'm wondering if it's worth it to the
>> Postgresql community if I dig further to see why those tests failed;
>> considering that obviously 7.4-dev fixed all of them.
>
> AFAIK there are not any portability enhancements in 7.4 that might
> explain such a difference. So the 7.3 results are interesting.  You
> should probably rule out pilot error though (does "select version()"
> confirm that you rebuilt with the newer compiler?)

template1=# select version();
                                   version
-----------------------------------------------------------------------------
 PostgreSQL 7.3.2 on alpha-unknown-netbsd1.6, compiled by GCC cc (GCC) 3.2.2
(1 row)

foo=# select version();
                                    version
--------------------------------------------------------------------------------
 PostgreSQL 7.4devel on alpha-unknown-netbsd1.6, compiled by GCC cc (GCC)
3.2.2
(1 row)


Thomas T. Thai