Discard stderr messages when using offline sniff

[roykuper13]

I'm opening this PR, because in my project i'm doing a lot of offline sniffing (mostly for filtering purposes). each time i'm using offline feature of sniff I get tcpdump's stderr, which is very very annoying to me, and for the user who uses my project.

I didn't make quiet an option in sniff for two reasons:

1. I don't think anyone would want to see this reading from file -, link-type EN10MB (Ethernet) message from tcpdump. it's not indicative and tells the user nothing. it's just very disturbing.
2. this PR affects only the offline sniff outputs, not other features of sniff, so putting quiet flag in sniff just doesn't make sense.

i'll be glad if you'll merge this, or to get suggestion for other solutions!
Thanks

[codecov]

Codecov Report

Merging #2558 into master will decrease coverage by 0.34%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #2558      +/-   ##
==========================================
- Coverage   88.28%   87.93%   -0.35%     
==========================================
  Files         243      246       +3     
  Lines       51006    52790    +1784     
==========================================
+ Hits        45032    46423    +1391     
- Misses       5974     6367     +393     

Impacted Files	Coverage Δ
scapy/sendrecv.py	85.99% <ø> (ø)
scapy/layers/tls/automaton_cli.py	67.23% <0.00%> (-16.54%)	⬇️
scapy/layers/tls/automaton_srv.py	66.39% <0.00%> (-11.77%)	⬇️
scapy/compat.py	74.69% <0.00%> (-3.51%)	⬇️
scapy/layers/tls/record_tls13.py	81.76% <0.00%> (-1.58%)	⬇️
scapy/arch/windows/__init__.py	70.58% <0.00%> (-1.55%)	⬇️
scapy/layers/netflow.py	88.51% <0.00%> (-0.82%)	⬇️
scapy/route.py	90.51% <0.00%> (-0.73%)	⬇️
scapy/main.py	72.80% <0.00%> (-0.41%)	⬇️
scapy/data.py	92.17% <0.00%> (-0.35%)	⬇️
... and 32 more

[roykuper13]

I'm not sure how debug the failure in travis pipline.
For example, in job #8670.2, All tests have FAILED=0 and in the end
___________________________________ summary ____________________________________ pypy-linux_root: commands succeeded codecov: commands succeeded congratulations :) ERROR: Too many parameters : [/tmp/tmp.b2K0h2nmPJ] The command "bash .config/ci/test.sh" exited with 1.

So i assume I have no failure, and there's a generic problem related to the travis.yml file?

[gpotter2]

Yes. We very recently migrated our test infrastructure, this issue is being tracked #2554

[guedou]

I am fine with defaulting to True for the tcpdump() quiet option, but I want to a have sniff() accept the quiet option as file. This will simplify debugging tcpdump errors. Thanks.

[roykuper13]

I am fine with defaulting to True for the tcpdump() quiet option, but I want to a have sniff() accept the quiet option as file. This will simplify debugging tcpdump errors. Thanks.

@guedou
alright, it's done, you can check my recent changes 👍

[guedou]

It looks good to me. @p-l- @gpotter2 what it is your call?

[gpotter2]

I fail to see what the use case of debug_file would be. Who would ever redirect it to a file ? This sounds like a totally unnecessary change, and breaks an API that has been around for three years 2a8b96a

I'd rather just have quiet a default

[p-l-]

An option to have tcpdump's stderr sent to stderr would be enough IMO

[roykuper13]

I fail to see what the use case of debug_file would be. Who would ever redirect it to a file ? This sounds like a totally unnecessary change, and breaks an API that has been around for three years 2a8b96a

I'd rather just have quiet a default

An option to have tcpdump's stderr sent to stderr would be enough IMO

@gpotter2 @p-l-
Agree with you guys. Added a quiet option to sniff, so now when set to True, tcpdump's stderr will be discarded 👍
Pushed my changes, so you can now review it.

[gpotter2]

IMO the best solution would be to always discard it except when the sub process returncode is 1.
This would need to pipe the stderr, retrieve during communicate(), and print it only on failure.

[roykuper13]

IMO the best solution would be to always discard it except when the sub process returncode is 1.
This would need to pipe the stderr to a buffer

@gpotter2 @p-l-
So i'm confused because you guys suggest different things.. what should I do then?
I quite support p-l-'s one to be honest.

[guedou]

I am fine with the current proposal. Relying on tcpdump is complicated as it does not have the same behavior across all platforms that we support.

[p-l-]

I am fine with the current proposal. Relying on tcpdump is complicated as it does not have the same behavior across all platforms that we support.

Totally, that's why I suggest to let tcpdump stderr go to scapy's stderr if the user wants so (by setting debug=True)

[roykuper13]

I am fine with the current proposal. Relying on tcpdump is complicated as it does not have the same behavior across all platforms that we support.

Totally, that's why I suggest to let tcpdump stderr go to scapy's stderr if the user wants so (by setting debug=True)

@guedou @p-l-
updated my changes, so now the user can "quiet" tcpdump's stderr if he wants. You can review it :)

[guedou]

Thanks!