[schema migration] fix backend compatibility issue by using type_from_constraints / sql_change_col_type instead of bare ALTER COLUMN

[schema migration] don't try to drop column if backend doesn't support column alteration (eg sqlite)

[session] when rollback_on_failure is false, we should not try to reconnect if the sql raised an error

Use SQL standard ALTER TABLE command (with ADD instead of ADD COLUMN)

reduced logging verbosity in ldapuser source

use sessionid in log message

constraints can contain unicode values

[entity] fix .complete so we don't crash on missing mandatory relation. Also simplify code. Closes #1058267

closes #1059729: cw-ctl upgrade : i18n upgrade not performed if DB does not need upgrade

[devtools/qunit] prefix jquery_dir with 'file://' (as this is used from within a web browser)

[devtools/qunit] Locate jquery through cubicwew config to handle installed setup

[rql2sql] use new dbhelper.sql_concat_string function to deal with sql string concatenation. This together with lgdb 1.0.4 fixes #1055133: SQL Server support : string concatenation issue

[session] update session's timestamp in session.execute, so long running transactions are not erroneously closed by the repository

[repo session] expected argument is the source, not the session

[devtools/qunit] in safe mode, ffox launches a stupid popup

[ldapuser] quieter ldap result info, fix commented block style

[qunit] use safe mode for firefox

[qunit] Move qunit.js and qunit.css in devtools/datadirectory

[test] fix stupd name error

[test] fix test w/ closed connection pb

[ms] orderby/groupby/having attributes of Select node should be resetted to empty tuple, not None

[devtools/teslib] be more defensive on rollbacks

[rql2sql] when using HAVING to by-pass rql limitation (not to filter on result of an aggregat function), we should emit SQL that doesn't use HAVING to avoid potential backend error because variables are not grouped. Closes #1061603.

[doc/book/faq] question on security

[qunit] check that the browser has not died by himself before we asked him to stop.

[qunit] improve output of suprocess error The exception message now include the stdout and stderr of the called process

[test] workaround connection leak making tests fail under windows environment (though we've still probably some open connection leak to resolve...)

[web request] in python 2.5 hashlib has no .hash attribute

[web actions] fix recent refactoring, leading to missing redirect parameters in action's url

[packaging] more cleanups, handle javascript test files