lose IN_CLOSE_WRITE event

[zhaochunqi]

Hi，
I've tried watch IN_CLOSE_WRITE event for my new files .
It works pretty good when I watch IN_CLOSE_WRITE only, when I watch IN_CREATE and IN_CLOSE_WRITE , it will lose lot's of IN_CLOSE_WRITE event .

It dosen't work when I import pyinotify either. I bugs me a lot.Hope you give me some advice.

[seb-m]

Hello,

Ok, what happens is you experience both this issue #2 and the effects of its mitigation.

Concretely when you copy and attempt to automatically watch a directory 01/ with multiples sub-directories there is a kind of concurrency issue between the moment files are copied and the moment when watches are added, that means for instance that if 01/01/* files are copied and then the 01/02/ directory is created, by the time Pyinotify has processed the events related to 01/01 there are some chances that 01/02 will already be created and its files copied. Because there is nothing we can really do to completely fix this issue, it was decided in #2 that the least bad behavior was to generate IN_CREATE for simulating creation of files contained in 01/02 because we know that this directory is new therefore its files are new, this is why you observe IN_CREATE events. However, others events are not simulated because we can't be sure that a file was either closed after a write or without a write for instance.

So in short, there is nothing we can't really do in this case and you can only rely on IN_CREATE events. Also if you really want to avoid this issue when you copy a directory, then you have to make sure it doesn't contain other directories.