I've updated the description to sum up what we discussed on the merge proposal, I hope it matches what you had in mind.
After further discussions, I think this change might be too big for stable, as it requires introducing a new ir.rule for ir.sequence, which does not exist now, and will absolutely require a data update - otherwise it will cause regressions in all existing multi-company setups which were not even affected by the original problem.
I suggest we keep the current behavior in 6.0, relying on the security rules of res.company when determining the sequences to use (in essence delegating the visibility of ir.sequence to res.company). It should not be a major problem for stable, provided that bug 809178 is fixed (I could not reproduce it, btw).
Then for 6.1, we can improve it as described above.
Hi Leonardo,
I've updated the description to sum up what we discussed on the merge proposal, I hope it matches what you had in mind.
After further discussions, I think this change might be too big for stable, as it requires introducing a new ir.rule for ir.sequence, which does not exist now, and will absolutely require a data update - otherwise it will cause regressions in all existing multi-company setups which were not even affected by the original problem.
I suggest we keep the current behavior in 6.0, relying on the security rules of res.company when determining the sequences to use (in essence delegating the visibility of ir.sequence to res.company). It should not be a major problem for stable, provided that bug 809178 is fixed (I could not reproduce it, btw).
Then for 6.1, we can improve it as described above.
What do you think?