Thank you for your prompt answer.

I installed FB 1.0.3 on a virtual machine. And tried to do a gbak. The error
message is the same :

gback: ERROR : BLOB not found
gback: ERROR : gds$_compile_request_failed

:-(



Tiberiu





From: Helen Borrie
Sent: Saturday, July 19, 2008 10:33 AM
To: firebird-support@yahoogroups.com
Subject: Re: [firebird-support] BLOB not found


At 05:08 PM 19/07/2008, you wrote:

>Hi all,
>
>Database "C:\DEVELOP\GESTOCW\gestoc.gdb"
>Database header page information:
> Flags 0
> Checksum 12345
> Generation 84877
> Page size 4096
> ODS version 10.0
> Oldest transaction 78660
> Oldest active 84866
> Oldest snapshot 84866
> Next transaction 84867
> Bumped transaction 1
> Sequence number 0
> Next attachment ID 50
> Implementation ID 16
> Shadow count 0
> Page buffers 0
> Next header page 0
> Database dialect 3
> Creation date Oct 23, 2002 14:49:10
> Attributes
>
> Variable header data:
> *END*
>
>Using Firebird 2.1.1
>
>When I open the database (with IBExpert) everything is fine, when I choose
>one table, I can see the metadata, but when I try to access the data from
>that table, I receive the next error message:
>
>"Unsuccessful execution caused by system error thet preclude successful
>execution of subsequent statements.
>BLOB not found"
>
>Receiving this when I try to backup (gbak) the database.
>
>What should I do ?

You have a database here that was created under IB6 or Firebird 1.0 in 2002.
If you plan to use this database with Fb 2 or higher, it is past time to do
a proper port of the database to use the version 11 and higher on-disk
structures.

The first thing I would try to do with this database is to install Firebird
1.0.3 on a spare machine and try to back up a copy of your database file
there. If that backup succeeds, I would then take it and restore it *safely*
in some spare disk space on the host machine where Fb 2.1.1 is running.

Then, if *that* succeeds, post an item in the tracker describing the whole
experience. It is just possible that 2.1.1 gbak implementation is expecting
some system table structure (a BLOB) that the very old ODS 10 databases
didn't have. Possibly some oversight in unit testing has given rise to this
unexpected behaviour when applications query the system tables. If so, my
guess is that it could relate to the new comment fields in the system
tables, which are text BLOBs of a new subtype.

./heLen