reset consumer

[firesWu]

Reset consumer every time when finish one layer.

If consumer not reset, may cause parent cache loading problem.

1. cur_parent_ptr may point to a released memory location, causing panic.
2. cur_parent_ptr may point to a wrong parent cache window, resulting wrong pc1 result.

[qy3u]

The bug you mentioned doesn't seem to be due to the consumer value.
The value of consumer has been correctly set to 1 here

[firesWu]

The bug you mentioned doesn't seem to be due to the consumer value. The value of consumer has been correctly set to 1 here

The thread that calls create_label_runner was created before consumer being reset to 1. So we cannot be sure that the consumer used here and here is already reset to 1. @qy3u

[qy3u]

I think you are right.
The consumer needs to be reset before create_label_runner.

[vmx]

I've only had a quick look, but to me it sounds like it is the right thing to do, to also reset the consumer.

[vmx]

@firesWu Sorry for the long delay. I know had a closer look. You change looks absolutely right. Thanks for spotting this issue. I was recently also deep into that part of the code, but I haven't spotted this issue. Good job!

Could you please rebase your change to current master (so that we don't end up with strange merge commits)?

[firesWu]

@vmx I have rebase my change to current master