Re: [HACKERS] Numeric type
От | jwieck@debis.com (Jan Wieck) |
---|---|
Тема | Re: [HACKERS] Numeric type |
Дата | |
Msg-id | m0zxvCV-000EBRC@orion.SAPserv.Hamburg.dsh.de обсуждение исходный текст |
Ответ на | Re: [HACKERS] Numeric type ("Thomas G. Lockhart" <lockhart@alumni.caltech.edu>) |
Список | pgsql-hackers |
> > > > OK, I give up :) How do I use the numeric type? > > Works here. Don't you hate when that happens: > > postgres=> create table n1 (n numeric(10,5), d decimal(10,5)); > CREATE > postgres=> insert into n1 values ('1.23456', '1.23456'); > ERROR: overflow on numeric > ABS(value) >= 10^0 for field with precision 2086 scale 53380 > > *sigh* > > Any hints on where to look? I did a CVSup update, then a "cvs -PdA" on > my development tree, a "make clean install" and an initdb. I did not try > a full clean checkout, but would think that it wouldn't help. Seems to me like the typmod value given to Numeric numeric(Numeric num, int dummy, int32 typmod) got mangled up somewhere. First you should check that the atttypmod in pg_attribute for n1.n and n1.d is correct. It should have the value ((precision << 16) | scale) + 4 and must be 655369 for a numeric(10,5). If that isn't the case, you might take a look at include/nodes/parsenodes.h. The typmod field of struct TypeName should be of int32 (I changed that shortly before moving numeric into the backend). If that isn't the error, it would be good if you're a little familiar with gdb. If so, set a breakpoint on function numeric() and try to find where on it's way from pg_attribute.atttypmod to the call to numeric() it got mangled up. > > Anyone else running on Linux/i686/libc5 having success? If so, I'll go > back to working on outer joins... > > - Tom > Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #======================================== jwieck@debis.com (Jan Wieck) #
В списке pgsql-hackers по дате отправления: