Closed GoogleCodeExporter closed 9 years ago
Additional note:
seems it does not crash if there are less then million records included in the
index. the same happens for the query creating temp table.
Original comment by krzyszto...@gmail.com
on 21 Apr 2011 at 12:43
Hi,
Sorry, I didn't see the bug report earlier (and code.google.com doesn't send
the notifications by default).
But I wasn't able to replicate the issue: (64 bit debian, PG 9.0.1 and 9.0.4
compiled with --block-size=16).
The regression test of q3c worked fine and the following small test worked too:
q3ctest=# CREATE TABLE tmp( ra double precision, dec double precision);
CREATE TABLE
q3ctest=# INSERT INTO tmp select random()*400-10, random()*200-100 from
generate_series(0,10000000);
INSERT 0 10000001
q3ctest=# CREATE INDEX q3c_idx on tmp(q3c_ang2ipix(ra,dec)) ;
CREATE INDEX
q3ctest=# VACUUM ANALYZE tmp;
q3ctest=# select * from tmp where q3c_radial_query(ra,dec,180,1,0.1);
ra | dec
------------------+-------------------
.....
q3ctest=# show block_size ;
block_size
------------
16384
(1 row)
If you still experience the issue(and interested in solving it), it would be
nice to get a gdb stacktrace and/or the the list of ras,decs on which index
creation fails
Original comment by koposov
on 10 Jun 2011 at 3:08
Hello,
Thanks for checking this.
I'll try to prepare the test case, I did not check what exact values caused the
exception so will have to scan the table in greater detail.
Cheers,
Krzysztof
Original comment by Krzyszto...@gmail.com
on 10 Jun 2011 at 5:34
Actually I just noticed google skipped additional information I posted after I
found out that very big values for ra,dec were triggering this error. Synthetic
creation of similarly looking values did not crash the backend, so I will have
to use real data to get the stacktrace.
Original comment by Krzyszto...@gmail.com
on 10 Jun 2011 at 6:01
Hi Krzysztof,
I think I fixed the issue in r122 .
Before the patch I was able to crash the backend by doing this:
select sum(q3c_ang2ipix(pow(random()*100,100*random()),random())) from
generate_series(0,1000);
Now it doesn't crash anymore. I will put the official release in a few days
time, but you can just grab the updated q3cube.c from the repository.
Thanks for reporting the issue.
Sergey
Original comment by koposov
on 10 Jun 2011 at 1:23
Just in case, I've released 1.4.11 version which fixes your issue.
Original comment by koposov
on 13 Jun 2011 at 4:02
Original issue reported on code.google.com by
krzyszto...@gmail.com
on 21 Apr 2011 at 12:10