Обсуждение: (Maybe) helpful data surrounding crashes
Hi developers! Hi Dave!
Testing pgAdmin v.1.8.2 (Feb 5 2008), rev: 7050, on WinXP; host: pg 8.2.6 on
Debian Etch
I have been having occasional crashes with pgAdmin 1.8.x. I still cannot pin
down the cause, it only happens like once on a busy day. They have something
in common: always when I click on an object in the object browser. This time,
after clicking on a schema (the first time since I connected to this
database, other databases of the same "server connection"-object already
open), while retrieving schema data.
Maybe it is of relevance that I have 12 different "server connection"-objects
in my object browser. Two of them were open before the crash.
Other programs do not normally crash on this machine.
This is the tail of my pgAdmin-Log before the crash:
-- pgAdmin log start --
(...)
2008-03-28 19:09:27 QUERY : Set query (86.59.118.251:5432): SELECT oid,
format_type(oid, typtypmod) AS typname FROM pg_type
2008-03-28 19:09:27 QUERY : Set query (86.59.118.251:5432): SELECT t.oid,
t.*, format_type(t.oid, null) AS alias, pg_get_userbyid(t.typowner) as
typeowner, e.typname as element, description, ct.oid AS taboid
FROM pg_type t
LEFT OUTER JOIN pg_type e ON e.oid=t.typelem
LEFT OUTER JOIN pg_class ct ON ct.oid=t.typrelid AND ct.relkind <> 'c'
LEFT OUTER JOIN pg_description des ON des.objoid=t.oid
WHERE t.typtype != 'd' AND t.typname NOT LIKE E'\\_%%' AND t.typnamespace =
153217::oid
AND ct.oid IS NULL
ORDER BY t.typname
2008-03-28 19:09:27 QUERY : Set query (86.59.118.251:5432): SELECT c.oid,
c.xmin, c.relname, pg_get_userbyid(c.relowner) AS viewowner, c.relacl,
description, pg_get_viewdef(c.oid, true) AS definition
FROM pg_class c
LEFT OUTER JOIN pg_description des ON (des.objoid=c.oid and des.objsubid=0)
WHERE ((c.relhasrules AND (EXISTS (
SELECT r.rulename FROM pg_rewrite r
WHERE ((r.ev_class = c.oid)
AND (bpchar(r.ev_type) = '1'::bpchar)) ))) OR (c.relkind =
'v'::char))
AND relnamespace = 153217::oid
ORDER BY relname
2008-03-28 19:09:27 STATUS : Retrieving Schema details... (0.23 secs)
-- pgAdmin log end --
BTW: 1 '%' is redundant here: ... t.typname NOT LIKE E'\\_%%' ...
Regards
Erwin
On Fri, Mar 28, 2008 at 9:21 PM, Erwin Brandstetter <brandstetter@falter.at> wrote: > Hi developers! Hi Dave! > > Testing pgAdmin v.1.8.2 (Feb 5 2008), rev: 7050, on WinXP; host: pg 8.2.6 on > Debian Etch > > I have been having occasional crashes with pgAdmin 1.8.x. I still cannot pin > down the cause, it only happens like once on a busy day. They have something > in common: always when I click on an object in the object browser. This time, > after clicking on a schema (the first time since I connected to this > database, other databases of the same "server connection"-object already > open), while retrieving schema data. > > Maybe it is of relevance that I have 12 different "server connection"-objects > in my object browser. Two of them were open before the crash. I've fixed the redundant %'s you mentioned, and fixed some other subtle bugs that could have caused crashes in some specific circumstances (got any objects with % in the name?). Not sure what else to try unless you want to try getting intimate with a debugger :-p -- Dave Page EnterpriseDB UK Ltd: http://www.enterprisedb.com PostgreSQL UK 2008 Conference: http://www.postgresql.org.uk
dpage@pgadmin.org wrote: > I've fixed the redundant %'s you mentioned, and fixed some other > subtle bugs that could have caused crashes in some specific > circumstances (got any objects with % in the name?). Not sure what > else to try unless you want to try getting intimate with a debugger > :-p Thanks for your attention to the details. I strictly stick to non-quoted-7Bit-ASCII-lowercase-letters-only-oldschool names, for I am afraid of ghosts. So, no, no % in my names. pgAdmin wouldn't get a chance to choke on % in comments or in the body of views or functions, would it? And even if I'd get all flirtatious with a debugger, which I don't have the time for right now, I'd still have to be able to reproduce the crash. Regards Erwin P.S.: I do remember a crash case involving a '%' being sent through a "fprint variant" (rev. 5932), though. Ah, the old days! Beer was cheap and bugs reproducible back then ...
On Tue, Apr 1, 2008 at 3:52 AM, Erwin Brandstetter <brandstetter@falter.at> wrote: > dpage@pgadmin.org wrote: > > I've fixed the redundant %'s you mentioned, and fixed some other > > subtle bugs that could have caused crashes in some specific > > circumstances (got any objects with % in the name?). Not sure what > > else to try unless you want to try getting intimate with a debugger > > :-p > > Thanks for your attention to the details. I strictly stick to > non-quoted-7Bit-ASCII-lowercase-letters-only-oldschool names, for I am > afraid of ghosts. So, no, no % in my names. pgAdmin wouldn't get a > chance to choke on % in comments or in the body of views or functions, > would it? No, shouldn't do. > And even if I'd get all flirtatious with a debugger, which I don't have > the time for right now, I'd still have to be able to reproduce the crash. when you have a debugger installed it should offer you the chance to debug as soon as it crashes. You'd just need the right debug symbols to hand. > P.S.: I do remember a crash case involving a '%' being sent through a > "fprint variant" (rev. 5932), though. Ah, the old days! Beer was cheap > and bugs reproducible back then ... Err, yeah. That's kinda what I fixed yesterday (I found a few more cases :-( ). -- Dave Page EnterpriseDB UK Ltd: http://www.enterprisedb.com PostgreSQL UK 2008 Conference: http://www.postgresql.org.uk