Fixes for various issues against 2015.12

[eugeneia]

Issue #666

Print meaningful error message when unable to create /var/run/snabb:

$ ./snabb
core/shm.lua:111: Unable to create /var/run/snabb: Permission denied
stack traceback:
    core/main.lua:126: in function <core/main.lua:124>
    [C]: in function 'assert'
    core/shm.lua:111: in function 'mkdir'
    core/shm.lua:84: in function 'map'
    core/counter.lua:48: in function 'open'
    core/app.lua:28: in main chunk
    [C]: in function 'require'
    core/main.lua:116: in function 'initialize'
    core/main.lua:35: in function <core/main.lua:32>
    [C]: in function 'xpcall'
    core/main.lua:155: in main chunk
    [C]: at 0x0044fa80
    [C]: in function 'pcall'
    core/startup.lua:1: in main chunk
    [C]: in function 'require'
    [string "require "core.startup""]:1: in main chunk

Issue #658

• Print useful error message when unable to open PCI device:

(When the target device of device_info does not exist)

$ sudo SNABB_PCI0="0000:05:00.0" SNABB_PCI1="0000:05:00.1" ./snabb snsh -t apps.intel.intel_app
selftest: intel_app
lib/hardware/pci.lua:38: No such device: 0000:05:00.0
stack traceback:
    core/main.lua:126: in function <core/main.lua:124>
    [C]: in function 'assert'
    lib/hardware/pci.lua:38: in function 'device_info'
    apps/intel/intel_app.lua:150: in function 'selftest'
    program/snsh/snsh.lua:29: in function <program/snsh/snsh.lua:29>
    core/lib.lua:405: in function 'dogetopt'
    program/snsh/snsh.lua:63: in function 'run'
    core/main.lua:58: in function <core/main.lua:32>
    [C]: in function 'xpcall'
    core/main.lua:155: in main chunk
    [C]: at 0x0044fa80
    [C]: in function 'pcall'
    core/startup.lua:1: in main chunk
    [C]: in function 'require'
    [string "require "core.startup""]:1: in main chunk

(When the selected device exists but is not suitable)

$ sudo SNABB_PCI0="0000:04:00.0" SNABB_PCI1="0000:04:00.1" ./snabb snsh -t apps.intel.intel_app
selftest: intel_app
SNABB_PCI_INTEL[0|1]/SNABB_PCI[0|1] not set or not suitable.

• Make packetblaster bail out if no suitable devices are selected:

$ sudo ./snabb packetblaster replay program/snabbnfv/test_fixtures/pcap/64.pcap 0000:05:00.0
program/packetblaster/packetblaster.lua:49: <PCI> matches no suitable devices.
stack traceback:
    core/main.lua:126: in function <core/main.lua:124>
    [C]: in function 'assert'
    program/packetblaster/packetblaster.lua:49: in function 'run'
    core/main.lua:58: in function <core/main.lua:32>
    [C]: in function 'xpcall'
    core/main.lua:155: in main chunk
    [C]: at 0x0044fa80
    [C]: in function 'pcall'
    core/startup.lua:1: in main chunk
    [C]: in function 'require'
    [string "require "core.startup""]:1: in main chunk

Issue #579

Make snabbnfv traffic warn when the configuration file does not exist
initially:

BINARY    2.0M snabb
$ sudo ./snabb snabbnfv traffic 0000:03:00.1 foobar baz
snabbnfv traffic starting
WARNING: File 'foobar' does not exist.
load: time: 1.00s  fps: 0         fpGbps: 0.000 fpb: 0   bpp: -    sleep: 100 us
[...]

[lukego]

Great! Merged onto next, sorry about the wait.

What do you think about a future policy of suppressing stacktraces for well-defined detected errors? (Showing them only for unexpected errors caused by bugs.) For example with a main.fatal(msg...) function to print and terminate. That would seem friendly to end-users.

[petebristow]

Suppressing stack traces for configuration/setup errors and the like would be great.

[eugeneia]

In that case I would like to retain the ability to always print stack traces (a debug option or similar). I understand that end users might be put off by stack traces but I think developers profit from seeing the guts no matter how trivial the error.