Following code, with foo having portal type Foo which is
not installed yet (thus it is broken) :
connection = self.getConnection(portal)
foo = connection.importFile(file_obj) => this does a savepoint
foo._p_changed = 1

Then next savepoint was raising :
PicklingError: Can't pickle <class 'erp5.portal_type.Foo'>: it's not the same object as erp5.portal_type.Foo )

This was because we had :
foo.__class__.mro() : (<class 'erp5.portal_type.Foo'>, <class 'ZODB.broken.PersistentBroken'>,
 <class 'ZODB.broken.Broken'>, <type 'persistent.Persistent'>, <class 'erp5.portal_type.Foo'>,
 <class 'Products.ERP5Type.dynamic.lazy_class.GhostBase'>,
 <class 'Products.ERP5Type.dynamic.lazy_class.ERP5BaseBroken'>, <class 'Products.ERP5Type.Base.Base'>, *TAIL)

and

erp5.portal_type.__getattribute__("Foo").mro() :
(<class 'erp5.portal_type.Foo'>, <class 'Products.ERP5Type.dynamic.lazy_class.ERP5BaseBroken'>,
  <class 'ZODB.broken.Broken'>, <class 'Products.ERP5Type.Base.Base'>, *TAIL)

Classes where different because new class with new inheritance was created by
ZODB.serialize.ObjecReader.load_persistent which does :

if issubclass(klass, Broken):
    # We got a broken class. We might need to make it
    # PersistentBroken
    if not issubclass(klass, broken.PersistentBroken):
        klass = broken.persistentBroken(klass)

Since ERP5BaseBroken is patched to mostly be like a PersistentBroken,
make sure we allow serialization to identify ERP5BaseBroken as a
PersistenBroken class.

erp5_discussion: make Discussion Post content fulltext searchable



See merge request !66

Renaming base category "journal" to "ledger"

Because "journal" is not French, and this base category can be used in different modules.

This merge request also provides a tested method in erp5_upgrader to update the objects using the "journal" category during the post-upgrade.

See merge request !54

A bug fix of erp5_core/Folder_delete



See merge request !65

Category Tool is very special for an ERP5, because it is part of the dynamic
generation of objects' properties. Setting up a property sheet on it will break
ERP5.

As the renaming feature is related to the upgrade, we will attach the constraint
on the Template Tool, as a way to group all the upgrade features at the same place

Submit action must be disabled while solver is being updated. End user sometimes click too fast that submit action fails due to lack of 'my_tested_property_list'.

Split each page into a separate gadget.
Stop using patched renderjs.

When a delivery was splitted, then the new delivery was also splitted,
and so on, we were generiting ids like "1_split_0_split_0_split_0_split_0...".
It ended having simulation having a path up to 240 characters on some projects
due to this (near 255 catalog limit).

For keeping some information on the split, only keep "_split_" one time in the
id. If we really need to know exactly from which simulation movement a new
splitted simulation movement is coming from, we should use something else than
id.

Now ERP5Site_reindex is not called during test setup, thus some tools
added before portal_activities should be reindexed explicitly.

Only portal_catalog.searchResults supports ignore_unknown_columns, it does not make sense at the level of Products.ZSQLCatalog.SQLCatalog.Query

Since we now have a transaction cache, changes in user assignements are
only taken into account in the next transaction.

It allows user to correctly select the start time and time zone.

It allows to create new DateTime with the same timezone.