[tor-qa] fix for some tbb test suite issues ?

[Gunes Acar]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Nicolas,

Here's the patch:
https://github.com/gunesacar/tbb-fp-tests/blob/master/always.patch

It removes the check for the "always" option and prevents
printing of the test names that are not enabled.

A new corner case was: when you run some tests without running
the tor_bootstrap, stop_tor (TorBootstrap.pm) gets called before
running start_tor in the first place.

Then it (stop_tor) tries to kill a non-existing Tor process and gives
the following error: "Can't kill a non-numeric process ID..."

So I added a check for the tor pid to fix this.

Hope that's all fine, please check the patch since I don't know Perl
at all.

Cheers,
Gunes

On 08/19/2014 02:51 AM, Nicolas Vigier wrote:
> On Mon, 18 Aug 2014, Gunes Acar wrote:
> 
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>> 
>> Hi,
>> 
>> Recently, I had some issues when running the TBB test suite. I
>> could submit a patch if you agree that the following can/should
>> be improved:
>> 
>> 1-) Make it possible to pass newlayout=0 from the command line
>> and add documentation for the "newlayout" option.
> 
> There is a --no-newlayout command line option: 
> https://people.torproject.org/~boklm/tbbtests-doc/usage.html
> 
> I noticed that I forgot to regenerate the html version of the doc
> after it was changed, so maybe it is the reason you didn't see it.
> 
>> 
>> The error messages I got due to the "newlayout" option was very
>> hard to trace.
> 
> Indeed, I remember having the same problem previously, and the
> error message in this case was not very helpful. It could be
> improved.
> 
>> 
>> 2-) The description for the "enable-tests" option says" "Only run
>> the list of tests selected"
>> 
>> This is not completely true since 'tor_bootstrap' will run
>> regardless of that option (managed by the property "always =>
>> 1".)
>> 
>> We can either fix the doc or ignore "always=1" when the
>> "enable-tests" option is set.
> 
> In the begining, all the tests depended on 'tor_bootstrap' being
> run, so I added the "always => 1" property to avoid listing it in
> 'enable-tests' everytime. Now that we have other tests that don't
> require 'tor_bootstrap', this property has become a little
> annoying, so maybe it could just be removed.
> 
> An other possible improvement could be to add support for
> dependencies between tests, so that when a test is enabled, its
> dependencies are enabled too. But just removing the 'always'
> property would probably be good enough for now.
> 
>> 
>> 3-) When the suite skips a test due to "enable-tests", it should
>> print out smt. like "Skipping the test..." or shouldn't print the
>> test name at all. Currently if prints the following even though
>> the tests were skipped:
> 
> This would be a good change.
> 
> 
> Thanks for reporting those problems. Patches for any of those
> issues are welcome. Let me know if you need some help.
> 
> Nicolas
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJT89wBAAoJEPb7JcMmVt4gqnAIALHwy3V6g/z6FWRmm5w5Depx
8bpyQXnCKHdMIOc1y1RAlRufbzEsemu0Hm78DR2OkFV/HhOKBXxYsiv4yP4uaOfN
hUv/cnXMAvDZ1zfn/xqsfb0YYznlTfvyy80tyUTD0v9IyXXB0u8mHvrd70LrSKFe
lB6B7CP5n+IkXpf7tOATh2w1OCb6GNYY2deZGto/D8Fviwdb4rnT93kfZX6m6UXX
RvpFuofGfikAGOPC4CLsWrR0XT/XqxM9CmrNTDbF/LUwqJIrQh1QOUZzZYWjSGhB
pCnJ6s0MCnKNoxe0ba2vSG3Q6+7uYcATL/S+BACv9GygxeWkhwbwzsd2oszz4lM=
=B3eu
-----END PGP SIGNATURE-----