Re: float4/float8/int64 passed by value with tsearch fixup

Alvaro Herrera írta:
> Zoltan Boszormenyi wrote:
>
>
>>>> So I think we really do need to offer that compile-time option.
>>>> Failing to do so will be tantamount to desupporting v0 functions
>>>> altogether, which I don't think we're prepared to do.
>>>>
>>>>
>>> I have asked the Cybertec guys for a patch.  Since it's basically a copy
>>> of the float8 change, it should be easy to do.
>>>
>> Here's the patch (against current CVS) with the required change.
>> Please review, it passed make check with both --enable and
>> --disable-float4-byval.
>>
>
> Does it pass "make installcheck" in contrib?  I'm worried about
>

It seems to pass, see below.

> btree_gist in particular.  Perhaps the change I introduced in the
> previous revision needs to be #ifdef'd out?
>

Both --enable and --disable-float4-byval produced only this regression,
but it seems to be a tuple order difference.

============= running regression test queries        ==============
test tsearch2             ... FAILED

cat tsearch2/regression.diffs:

*** ./expected/tsearch2.out     Tue Nov 20 05:23:10 2007
--- ./results/tsearch2.out      Sat Apr 19 20:48:16 2008
***************
*** 1490,1497 ****
   w0        |   29 |     29
   y9        |   29 |     29
   zm        |   29 |     29
-  zs        |   29 |     29
   zy        |   29 |     29
   ax        |   28 |     28
   cd        |   28 |     28
   dj        |   28 |     28
--- 1490,1497 ----
   w0        |   29 |     29
   y9        |   29 |     29
   zm        |   29 |     29
   zy        |   29 |     29
+  zs        |   29 |     29
   ax        |   28 |     28
   cd        |   28 |     28
   dj        |   28 |     28

======================================================================




--
----------------------------------
Zoltán Böszörményi
Cybertec Schönig & Schönig GmbH
http://www.postgresql.at/