Hello,
looking at changes in the database structure from 3.2 to 3.3, I noticed the columns in active_watchers/rls_watches that have no default value and cannot be NULL.
The usual issue with this kind of upgrades is that existing rows have to be removed. active_watchers is with temporary records, and I guess that is the same with rls_watchers. No persistent data is to be lost, right?
Another one is extra_hdrs in silo table. I think it should be allowed to be NULL or default '' (empty string), so stored messages will not be lost - the check over the code showed inserting '' if no extra headers are present at message storage time.
Cheers, Daniel
Hi Daniel,
rls_watchers is the same as active_watchers (holding state for presence dialogs). No persistent data is held in these tables.
Hugh (who did the extra headers development) is looking at the silo table.
Regards,
Peter
On Fri, 2012-05-04 at 09:40 +0200, Daniel-Constantin Mierla wrote:
Hello,
looking at changes in the database structure from 3.2 to 3.3, I noticed the columns in active_watchers/rls_watches that have no default value and cannot be NULL.
The usual issue with this kind of upgrades is that existing rows have to be removed. active_watchers is with temporary records, and I guess that is the same with rls_watchers. No persistent data is to be lost, right?
Another one is extra_hdrs in silo table. I think it should be allowed to be NULL or default '' (empty string), so stored messages will not be lost - the check over the code showed inserting '' if no extra headers are present at message storage time.
Cheers, Daniel
Hello, The current definition of the extra_hdrs is: <column> <name>extra_hdrs</name> <type>text</type> <default/> <description>Extra headers that must be restored</description> </column> which results in extra_hdrs TEXT DEFAULT '' NOT NULL in mysql and postgres. This has a default value. Do you think it should have both a default and be allowed to be NULL?
Regards, Hugh
On 04/05/2012 08:40, Daniel-Constantin Mierla wrote:
Hello,
looking at changes in the database structure from 3.2 to 3.3, I noticed the columns in active_watchers/rls_watches that have no default value and cannot be NULL.
The usual issue with this kind of upgrades is that existing rows have to be removed. active_watchers is with temporary records, and I guess that is the same with rls_watchers. No persistent data is to be lost, right?
Another one is extra_hdrs in silo table. I think it should be allowed to be NULL or default '' (empty string), so stored messages will not be lost - the check over the code showed inserting '' if no extra headers are present at message storage time.
Cheers, Daniel
Hello,
you are right.
The tool I used to compare the databases structures and get the SQL statements for upgrade seems to be buggy for text columns, it gave the result:
ALTER TABLE silo ADD COLUMN extra_hdrs text NOT NULL;
But in kamctl the definition is with DEFAULT ''
Thanks, Daniel
On 5/4/12 11:57 AM, Hugh Waite wrote:
Hello, The current definition of the extra_hdrs is:
<column> <name>extra_hdrs</name> <type>text</type> <default/> <description>Extra headers that must be restored</description> </column> which results in extra_hdrs TEXT DEFAULT '' NOT NULL in mysql and postgres. This has a default value. Do you think it should have both a default and be allowed to be NULL?
Regards, Hugh
On 04/05/2012 08:40, Daniel-Constantin Mierla wrote:
Hello,
looking at changes in the database structure from 3.2 to 3.3, I noticed the columns in active_watchers/rls_watches that have no default value and cannot be NULL.
The usual issue with this kind of upgrades is that existing rows have to be removed. active_watchers is with temporary records, and I guess that is the same with rls_watchers. No persistent data is to be lost, right?
Another one is extra_hdrs in silo table. I think it should be allowed to be NULL or default '' (empty string), so stored messages will not be lost - the check over the code showed inserting '' if no extra headers are present at message storage time.
Cheers, Daniel
-- Hugh Waite Senior Design Engineer Crocodile RCS Ltd.
-
Daniel-Constantin Mierla -
Hugh Waite -
Peter Dunkley