We need to decide what to do when the socket is closed but < the expected number of bytes is received.

I don't have much input for this, just wanted to note that there's two more places in our code base with a socket.recv:

• clients/neic/client.py line 204
• clients/seedlink/client/seedlinkconnection.py line 1202

Also, this might even be considered a bug fix? (Which would qualify it for maintenance_1.0.x and thus 1.0.3?)

+TESTS:clients.earthworm,clients.seedlink (to make Travis run those network modules' tests as well)

@megies: I checked those clients (neic, seedlinkconnection) but it appeared they are using nonblocking sockets, which should generate an exception instead of fail silently.

When I put in the pr, I checked but didn't find the maintenance branch. Let me know if you need me to update the pr.

When I put in the pr, I checked but didn't find the maintenance branch. Let me know if you need me to update the pr.

OK. Also, there should be a tick box for you on this page to the right "Allow edits from maintainers. ". If you tick it we should be able to push to your branch if necessary.

@megies: I checked those clients (neic, seedlinkconnection) but it appeared they are using nonblocking sockets, which should generate an exception instead of fail silently.

Alright. Like I said, I'm no expert on sockets so I'll leave this to @QuLogic and @krischer and whoever feels qualified to review this.

Looks good to me and I don't think the change can have any negative side effects! Thanks a lot!

I guess the memory leak happened because it kept on looping and adding empty things to the chunks list?

Can you please make your changes to the maintenance_1.0.x branch and force push? Let us know if you need help with this.

We need to decide what to do when the socket is closed but < the expected number of bytes is received.

It will probably just raise some kind of parsing error later down the line which is fine IMHO.

Rebasing ended up being a bad idea... Changes updated to start from maintenance_1.0.x.

Yea rebasing does not work here. You could have cherry picked the commits on top of a branch off of maintenance. But whatever you did - looks good now :-)

Can be merged depending on what CI says.

@megies: Are the seedlink and earthworm tests supposed to be executed everywhere? I don't see them anywhere.

@krischer, no, just Travis picks that up currently (see #1408 (comment)). My Windows-fu wasn't up to adding it to Appveyor too, and the docker scripts are pretty messy already, so I didn't want to add more mess on top.

Now, don't ask me why Travis didn't run this PR.. (maybe it got confused by the rebase)

I see. I'm merging in any case - I'm fairly sure this works based on the changes + the socket module documentation. Additionally I tested it locally for Python 2 and 3.

@jmfee-usgs: Thanks for this!

I ran this PR through our docker images, looking good. 👍 http://tests.obspy.org/?git=701bdb68fd&node=docker-

Thanks @jmfee-usgs!