ENH add ResourceTracker to improve ressource clean up #204
Conversation
|
French typo everywhere: ressource => resource :) |
|
A batch of comments. Overall this LGTM as well (there are duplicated comments with the |
|
I need to somehow rebase with #202. |
ac477db
to
2cbbc88
|
A first batch of comments. This is going in the right direction :) |
d9af7d9
to
cc754f7
|
@tomMoral I addressed most of the comments from your previous review. There is one thing though apparently, Recent support has been added to enable files descriptors to child processes in windows, (post windows 7, and post python3.7). Should we dive further into it? Or stick to POSIX-only support? |
36da509
to
88eeccc
|
A (large) batch of comments, many are nitpicks. This looks very good, thanks @pierreglaser for the hard work! I will do a second pass in the afternoon if possible. |
|
OK @tomMoral I addressed most of your comments. I am somewhat hesitant about rewriting the |
Co-Authored-By: Thomas Moreau <thomas.moreau.2010@gmail.com>
115554f
to
36092f9
I understand but we already have some slight differences so I think it would be fine. The question is more whether it is useful or not. I think it allows you to avoid having an empty set |
|
LGTM! Nice work @pierreglaser |
This PR aims at extending the current
SemaphoreTrackerto make it able to track more generic named resources, such as folders, and potentiallymemmapand shared memory.Many small changes in files because of the name refactoring. The main change is the registering/unregistering API extension: when one wants to track a new resource, one must (along with the resource name), specify the resource type. This is necessary for the
RessourceTrackerto know which cleanup function to use for each resource before shutdown.Base branch: https://github.com/tomMoral/loky/pull/202