[Erp5-dev] there is code destroying data into ERP5 currently

[Sébastien Robin]

Hi,

There is code destroying data into ERP5Type/Core/Folder.py

You have been warned, revisions upper than 16294 WILL DESTROY DATA.

We will fix it today.

Seb.

[bartek]

Post by Sébastien Robin
Hi,
There is code destroying data into ERP5Type/Core/Folder.py
You have been warned, revisions upper than 16294 WILL DESTROY DATA.
We will fix it today.

Is it already fixed? Is it safe now to upgrade? Basically, what is the
status of HEAD now - is it more or less usable, or does it still contain
some ultra-experimental stuff?

Bartek

Post by Sébastien Robin
Seb.

[Łukasz Nowak]

Hello,

On 2007-09-18, 11:47:31

Post by bartek
Is it already fixed? Is it safe now to upgrade? Basically, what is
the status of HEAD now - is it more or less usable, or does it still
contain some ultra-experimental stuff?

I'm currently upgrading copy of our production system to rev 16456 and
for now no object is disappearing and no problems spotted.

BTW - wouldn't be good to create HowToUpgrade page on wiki? To point
problems with workflows, business template overriding and create some
kind of "good upgrade scenario"?

Regards,
Luke

[Jean-Paul Smets]

Post by bartek

Post by Sébastien Robin
Hi,
There is code destroying data into ERP5Type/Core/Folder.py
You have been warned, revisions upper than 16294 WILL DESTROY DATA.
We will fix it today.

Is it already fixed? Is it safe now to upgrade? Basically, what is the

It should be. About 10 developers currently use the head branch.

Post by bartek
status of HEAD now - is it more or less usable, or does it still contain
some ultra-experimental stuff?

HEAD does not contain ultra-experimental code or does not use
experimental code (we keep such code away). The incident of last week
happened although the code which was committed had been unit tested and
atlhough we had tested upgrade (but not enough) before committing. It
was a complex situation. We are really sorry.

The nice thing with Zope is that it is always possible to revert
database transactions and recover lost data.

Regards,

JPS.

Post by bartek
Bartek

Post by Sébastien Robin
Seb.

[Pelletier Vincent]

Post by Sébastien Robin
There is code destroying data into ERP5Type/Core/Folder.py

BTW I *think* the code did not actually destroy data.

What it did was to store a folder-handling plugin instance on the folder when
folder instance gets modified.
As long as this plugin instance is not there, folder behaves like usually, but
as soon as it's there its methods are called... on the plugin instance.

Which means that instead of searching for subobjects on the Folder (where they
still are) it searched for them on the plugin instance, where there is none
(actually, there is just the one just added).

I have *NOT* checked this assumption, but it popped into my mind when I took a
look at Sebastien's fix and comparing to original code.

If this is true, reverting the Folder code to previous version should make
seemingly-lost objects pop up again, but the downside is the apparent (again)
disapearance of all objects created with the new code...

[Yusei TAHARA]

Hi.

Post by Pelletier Vincent
If this is true, reverting the Folder code to previous version should make
seemingly-lost objects pop up again, but the downside is the apparent (again)
disapearance of all objects created with the new code...

No, lost objects does not pop up again. I already try this approach last week,
and failed. I read the code. it does clear btree object so that objects is cleared.
but, of course, we can find lost objects from Data.fs. it keeps old transactions,
even if you cannot undo from ZMI.