Re: [QUESTIONS] regression failed for numerology. - Mailing list pgsql-hackers

From Thomas G. Lockhart
Subject Re: [QUESTIONS] regression failed for numerology.
Date
Msg-id 35250153.A87B3B39@alumni.caltech.edu
Whole thread Raw
List pgsql-hackers
> >  regression test failed for numerology with 6.3.1 (compiled with
> >MB=EUC_JP option by gcc-2.7.2) on x86 Linux-2.0.33.
> >There was not such an error with 6.3.  Is anyone know about this?
> test=> select * from f where (f > -2147483647);
> ERROR:  There is no operator '>' for types 'float8' and 'int4'
> Look like "> plus_number" and  ">= minus_number" are ok, only
> "> minus_number" fails.

Yes, it appears to be a side-effect of a change I made to ensure that
all "unary minus" math expressions are handled correctly. If you revert
scan.l/scan.c to those from v6.3 this particular problem should
disappear (but the original reported problem, which is apparently not
commonly seen, will reappear).

We should have caught this before releasing v6.3.1 :(

I am working on automatic type conversion capabilities for v6.4, and
would expect that this problem goes away.

                    - Tom

pgsql-hackers by date:

Previous
From: "Edwin S. Ramirez"
Date:
Subject: Problem with 6.3.1 under Linux RH5.0
Next
From: "Thomas G. Lockhart"
Date:
Subject: Re: [PATCHES] Re: pg_dump -z [Please don't apply this]