Re: Testing v1.7: Server restart/crash crashes pgAdmin - Mailing list pgadmin-hackers

From Leszek Trenkner
Subject Re: Testing v1.7: Server restart/crash crashes pgAdmin
Date
Msg-id c54a8$4654da82$594f4043$24425@news.chello.pl
Whole thread Raw
Responses Re: Testing v1.7: Server restart/crash crashes pgAdmin
List pgadmin-hackers
Hello,

I recently run into problem with query analyze on postgres 8.3 devel
(repetable crash of server, yet to recreate exact conditions), and it
also caused my pgAdmin 1.7.0 (6301M) on Linux (Ubuntu Fiesty 7.04) to
crash/hang/name it.

Steps to recreate :

1. Connect to server, which you'll be able to restart.
2. Open query window for this server.
2a. Run any query (not sure if it's really needed).
3. Restart server.
4. Try to run any query. (Nothing really happens?!)
5. Try to run any query again. Window pops up saying "An error has
occured: no conection to server" which seems fine, but...

At this point dialog box does not respond to clicks. Main window, and
dialog box are not refreshed any more. None of them respond to user
action. The only way out is to kill pgAdmin process.

I think there is some old server-disconnection recovery code involved,
AFAIR in 1.2 (and 1.4 maybe?) server disconnection crashed pgAdmin when
using browser tree, but in 1.6 and 1.7 gracefully(?:-)) recovers noticing
user in a dialog boxes that the connection to server was lost.

Note:
There can be a few (up to 3 I think) of them saying "An error has
occurred:
FATAL:  terminating connection due to administrator command
server closed the connection unexpectedly
    This probably means the server terminated abnormally
    before or while processing the request.
".
Last one is different and asks "Close server browser? If you abort, the
object browser will not show accurate data."

Maybe the same code can be used to protect query windows from server
disconnection issues?

--
Greetings,
Leszek Trenkner

pgadmin-hackers by date:

Previous
From: Leszek Trenkner
Date:
Subject: Re: Testing v1.7: Server restart/crash crashes pgAdmin
Next
From: Leszek Trenkner
Date:
Subject: Testing v1.7: Reverse engineering of multiparameter aggregate functions fails