Unreproducible bug in snapshot import code
Hi All,
I have not been able to reproduce this error, but wanted to report this
in case it might be useful.
Commit -id: 0f39d5050dc0dce99258381f33f1832c437aff85
Configure options:
--prefix=/mnt/storage/gurjeet/dev/builds//master/db --enable-debug
--enable-cassert CFLAGS=-O0 --enable-depend --enable-thread-safety
--with-openssl
Used VPATH to build postgres.
Attached files: terminal1.txt termonal2.txt config.log
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACK
As you can see the SET TRANSACTION SNAPSHOT was the first statement in
that transaction, and yet the ERROR message says that it is not. That
snapshot id was generated in another session (terminal2.txt), and was
generated outside any transaction; that's the only peculiar thing I can say
about that snapshot-id.
As is evident from the rest of the lines in terminal1.txt, I have not
been able to reproduce this error again; I tried even with a clean build of
the sources.
I have taken a look at the code and everything tells me that the error
should have been:
ERROR: a snapshot-importing transaction must have isolation level
SERIALIZABLE or REPEATABLE READ
Maybe it was not a clean build the first time, and the code may have
wrongly linked with previously compiled .o files.
Regards,
--
Gurjeet Singh
EnterpriseDB Corporation
The Enterprise PostgreSQL Company
Gurjeet Singh <singh.gurjeet@gmail.com> writes:
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACK
As is evident from the rest of the lines in terminal1.txt, I have not
been able to reproduce this error again; I tried even with a clean build of
the sources.
Given the weird capitalization of the BEGIN command, I'm guessing that
you used tab-completion to enter it. I wonder if this could have been
affected by the execution of some query in support of tab completion?
I couldn't reproduce it either on the basis of that guess, though.
regards, tom lane
Excerpts from Gurjeet Singh's message of lun oct 24 13:55:44 -0300 2011:
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACKAs you can see the SET TRANSACTION SNAPSHOT was the first statement in
that transaction, and yet the ERROR message says that it is not.
Maybe the tab-completion feature issued a query before the set
transaction command.
--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support
On Mon, Oct 24, 2011 at 1:08 PM, Alvaro Herrera
<alvherre@commandprompt.com>wrote:
Excerpts from Gurjeet Singh's message of lun oct 24 13:55:44 -0300 2011:
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACKAs you can see the SET TRANSACTION SNAPSHOT was the first statement
in
that transaction, and yet the ERROR message says that it is not.
Maybe the tab-completion feature issued a query before the set
transaction command.
I have tried reproducing the bug starting from 1 and 2 transactions before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.
Regards,
--
Gurjeet Singh
EnterpriseDB Corporation
The Enterprise PostgreSQL Company
On Tue, Oct 25, 2011 at 2:41 AM, Gurjeet Singh <singh.gurjeet@gmail.com> wrote:
On Mon, Oct 24, 2011 at 1:08 PM, Alvaro Herrera <alvherre@commandprompt.com>
wrote:Excerpts from Gurjeet Singh's message of lun oct 24 13:55:44 -0300 2011:
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACKAs you can see the SET TRANSACTION SNAPSHOT was the first statement
in
that transaction, and yet the ERROR message says that it is not.Maybe the tab-completion feature issued a query before the set
transaction command.I have tried reproducing the bug starting from 1 and 2 transactions before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.
I could reproduce it when I typed TAB just after typing "set" in "set
transaction snapshot".
As Tom and Alvaro pointed out, the tab-completion issues a query and which
prevents the "set transaction snapshot" command.
Regards,
--
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center
On Tue, Oct 25, 2011 at 2:33 AM, Fujii Masao <masao.fujii@gmail.com> wrote:
On Tue, Oct 25, 2011 at 2:41 AM, Gurjeet Singh <singh.gurjeet@gmail.com>
wrote:On Mon, Oct 24, 2011 at 1:08 PM, Alvaro Herrera <
alvherre@commandprompt.com>
wrote:
Excerpts from Gurjeet Singh's message of lun oct 24 13:55:44 -0300 2011
:
Starting from line 89 of terminal1.txt we see this snippet:
postgres=# rollback;
ROLLBACK
postgres=# begin TRANSACTION ;
BEGIN
postgres=# set transaction snapshot '000002C8-1';
ERROR: SET TRANSACTION SNAPSHOT must be called before any query
postgres=# rollback;
ROLLBACKAs you can see the SET TRANSACTION SNAPSHOT was the first
statement
in
that transaction, and yet the ERROR message says that it is not.Maybe the tab-completion feature issued a query before the set
transaction command.I have tried reproducing the bug starting from 1 and 2 transactions
before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.I could reproduce it when I typed TAB just after typing "set" in "set
transaction snapshot".
As Tom and Alvaro pointed out, the tab-completion issues a query and which
prevents the "set transaction snapshot" command.
Great! That settles it then. Reproducible, but not a bug.
Thanks,
--
Gurjeet Singh
EnterpriseDB Corporation
The Enterprise PostgreSQL Company
Gurjeet Singh wrote:
I have tried reproducing the bug starting from 1 and 2 transactions
before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.I could reproduce it when I typed TAB just after typing "set" in "set
transaction snapshot".
As Tom and Alvaro pointed out, the tab-completion issues a query and which
prevents the "set transaction snapshot" command.Great! That settles it then. Reproducible, but not a bug.
Yes, it is only tabs that query the database for completion that cause
this. Should this be documented somehow? (No idea how.)
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
On Fri, Oct 28, 2011 at 10:11 AM, Bruce Momjian <bruce@momjian.us> wrote:
Gurjeet Singh wrote:
I have tried reproducing the bug starting from 1 and 2 transactions
before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.I could reproduce it when I typed TAB just after typing "set" in "set
transaction snapshot".
As Tom and Alvaro pointed out, the tab-completion issues a query andwhich
prevents the "set transaction snapshot" command.
Great! That settles it then. Reproducible, but not a bug.
Yes, it is only tabs that query the database for completion that cause
this. Should this be documented somehow? (No idea how.)
If we have a doc section on psql's tab-completion, I think this needs to go
there. A note like:
"Trying to tab-complete on psql may send queries to the server, and
depending on the transaction state, execution of these queries may lead to
non-default/unexpected behaviour by the queries executed after
tab-completion. For example, ..."
Regards,
--
Gurjeet Singh
EnterpriseDB Corporation
The Enterprise PostgreSQL Company
On Fri, Oct 28, 2011 at 11:07:25AM -0400, Gurjeet Singh wrote:
On Fri, Oct 28, 2011 at 10:11 AM, Bruce Momjian <bruce@momjian.us> wrote:
Gurjeet Singh wrote:
I have tried reproducing the bug starting from 1 and 2 transactions
before
the one shown in snippet, and I used tab-completion to get the same
screen-output as termonal1.txt and yet it's not reproducible.I could reproduce it when I typed TAB just after typing "set" in "set
transaction snapshot".
As Tom and Alvaro pointed out, the tab-completion issues a query andwhich
prevents the "set transaction snapshot" command.
Great! That settles it then. Reproducible, but not a bug.
Yes, it is only tabs that query the database for completion that cause
this. Should this be documented somehow? (No idea how.)If we have a doc section on psql's tab-completion, I think this needs to go
there. A note like:"Trying to tab-complete on psql may send queries to the server, and depending
on the transaction state, execution of these queries may lead to non-default/
unexpected behaviour by the queries executed after tab-completion. For example,
..."
I have added the attached patch to document this limitation.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
Attachments:
psql.difftext/x-diff; charset=us-asciiDownload
diff --git a/doc/src/sgml/ref/psql-ref.sgml b/doc/src/sgml/ref/psql-ref.sgml
new file mode 100644
index ebb0ad4..a47af51
*** a/doc/src/sgml/ref/psql-ref.sgml
--- b/doc/src/sgml/ref/psql-ref.sgml
*************** testdb=> \set PROMPT1 '%[%033[1;33;40
*** 3272,3278 ****
exits and is reloaded when
<application>psql</application> starts up. Tab-completion is also
supported, although the completion logic makes no claim to be an
! <acronym>SQL</acronym> parser. If for some reason you do not like the tab completion, you
can turn it off by putting this in a file named
<filename>.inputrc</filename> in your home directory:
<programlisting>
--- 3272,3281 ----
exits and is reloaded when
<application>psql</application> starts up. Tab-completion is also
supported, although the completion logic makes no claim to be an
! <acronym>SQL</acronym> parser. The queries generated by tab-completion
! can also interfere with other SQL commands, e.g. <literal>SET
! TRANSACTION ISOLATION LEVEL</>.
! If for some reason you do not like the tab completion, you
can turn it off by putting this in a file named
<filename>.inputrc</filename> in your home directory:
<programlisting>