otherwise we have no % expansions

was too fast in previous attempt to fix the issue

regression introduced with commit cacea29b

Inventory API has minimal catalog table usage (join on uid  only), which
makes on-deletion catalog changes insufficient to make movements
disappear transactionally from Inventory API results.
As it may be expected that a set of movements deleted during a single
transaction are atomically removed from Inventory API results, also
change stock table.

If an object has no uid yet and has no acquisition context to the portal,
getUid() fails.

Default is derived from id(aq_base(ob)) which is even more efficient.

The method on XMLObject was also redundant because it inherited the same one
from Base.

(cherry picked from commit 347804af)

Because on DVCS with branches, history is not always linear, we must stop
using increasing integer for revisions because they cause too many conflicts
and a higher revision number does not necessarily means the a BT newer.
They're now a cryptographic hash of all the contents of the Business Template,
and they're computed automatically:
- by genbt5list
- at download
- at building (by default)
- at export

Add test to simulate the issue with the method TemplateTool.installBusinessTemplateListFromRepository

Using the method try to install Business Templates that are already
installed, the Business Templates is uploaded on ERP5 but not used. It
creates a lot of garbage because in the end you have many bt5s not
installed.

(cherry picked from commit 28e0f0a2)

Considering erp5_configurator_ung depending on erp5_configurator which in turn
depends on erp5_workflow, if erp5_workflow and erp5_configurator_ung are
specifically added to the list of bt5 installed with depends, then erp5_workflow
ended up being installed *after* erp5_configurator.

(cherry picked from commit 0101c3a0)

- stop using chdir
- make logging to stdout/stderr optional
- generateInformation: return a StringIO instead of writing to a temporary file

(cherry picked from commit ad309c80)

This reverts commit bf003551.
This commit will be re-applied along with necessary unit test changes.

Real-world users may copy-paste logins without realising they also copy
whitespaces.

_searchAndActivate typically calls callMethodOnObjectList with paths as
parameter. The result is that each listed document is loaded, which may
not be required or desired.
select_method_id avoids this loading by calling method_id once per
packet_size with a list of values extracted from catalog result as (first)
argument.

From time to time, it happened that after hitting button "run live test" the
result was not displayed even though the test was running. This hack should
clearly reduce the risk. If there is still issues from time to time, we will need
to improve design.

also fix some more columns showing category ids / not translated titles