fix(core-processor): fix empty allocations #3894
Conversation
grishasobol
requested review from
breathx,
gshep and
ark0f
and removed request for
breathx
There was a problem hiding this comment.
Could you please provide more detailed explanation how and why it was before and how will be from now on? @grishasobol
It was because there was no difference between empty allocations and completely freeing allocation. So, because core-processor updates allocation only in case they are not empty - we have a problem that after freeing all allocations we have no updates in storage after that. Now update will be. This is not very important bug. Only in case some program allocates lots of pages in one executions and then frees all of them in another execution, then allocations won't be freed, so that can cause problems for programs who wants to be sure that freed allocations can be accessible to be allocated in next executions. |
Actually I do not why this bug was in our code. Very old bug I think |
Resolves #3853
The behaviour of syscalls
allocandfreewill be changed a little. Now it will works correctly, so that all freed pages are removed from storage with their data as well.I don't think this is important change for gear programs already uploaded to our web3 networks, actually this bug could not be used for some kind of abuse and did not brings logical issues for gear programs.