enabled arclink decryption during download

[barsch]

lesser memory footprint while having the same execution speed as before

[barsch]

coverage/coveralls lies - it includes files which were not changed in this PR - however I really love the idea of this feature if it would work ...

[krischer]

Yea coveralls is funny sometimes...no clue why it lists untouched files.

The PR itself looks fine to me. I assume you locally tested on py2 and py3?

[barsch]

tests were fine for windows - still would be nice if someone would test it manually on linux

[krischer]

I get lots of timeouts when running test_decrypt.py but can run the normal arclinks tests just fine. Is this expected? Is encrypted data send over a different port?

[barsch]

hmm not good - I'll check at home if this is still running for me and report back - thanks for testing

[barsch]

oh and port 36000 needs to be open - see https://github.com/obspy/obspy/blob/master/obspy/clients/arclink/tests/test_decrypt.py#L35

[krischer]

I see. I'll test at home.

[krischer]

FYI: I just pushed a commit to this branch that allows to run the crypto tests with either supported crypto library.

[barsch]

careful: EVPError is only defined for M2Crypto ! -> unfortunatly as far I tested pycrypto does not raise anything if there is an error during decryption - so we have to stick with M2Crypto as default and only take pyCrypto as fallback

[krischer]

Hmm...there must be a way to figure out if the decrypt was successful, right? How does m2crypto even know? I mean we could always check the resulting miniseed file but m2crypto clearly does something else.

[barsch]

while using pycrypto it just fails later - either because it can't unpack using bzip2 or it can't read as its not a valid mseed

AFAIK m2crypto checks in the final call https://github.com/obspy/obspy/blob/master/obspy/clients/arclink/decrypt.py#L66 - so we could look into that

[megies]

+TESTS:clients.arclink

Travis should run ArcLink test suites for future commits/pushes to this branch now (see here). Please ping me if it doesn't work..

[megies]

What's the status here, @barsch @krischer?

[barsch]

the later commit from @krischer does imho not work - see comment above - so either we stick with my orginal implementation, leave it as it is or find something better

and triggering a new test run did not work, or?

[megies]

and triggering a new test run did not work, or?

I rebased and force-pushed right now, that will trigger CI..

[megies]

Current Travis commit status includes Arclink tests: https://travis-ci.org/obspy/obspy/builds/176344138

Not sure if tests cover decryption, though?

[megies]

I'll bump this to 1.2.0. Feel free to add to 1.1.0 if it gets ready in the next couple of weeks.

[barsch]

imho ready to merge

[megies]

Yep, CI looks good (we really should figure out why we have these erratic arclink "UNSET" problems.. or well.. once the VCR stuff is ready we're rid of it I guess)

Rebased on current master, mainly to get rid of some merge commits within this branch..

[barsch]

failing test is not related - merging

[megies]

@barsch this PR should also have modified arclink extra/optional dependencies.. anybody know how to modify a set of OR dependencies in setup.py? I couldn't find it..

https://github.com/obspy/obspy/blob/master/setup.py#L116

Should be semantically be something like..

 'arclink': ['m2crypto | pycrypto | cryptography'],

[barsch]

I don't think there is such option - I would take it out completely and just document it

btw: pycryptodome does also work - its an drop-in replacement for pycrypto