Re: [Boost-testing] Trunk testing broken.

Subject: Re: [Boost-testing] Trunk testing broken.
From: Belcourt, Kenneth (kbelco_at_[hidden])
Date: 2010-04-04 19:21:58

On Apr 4, 2010, at 10:15 AM, Vladimir Prus wrote:

> Rene Rivera wrote:
>
>> K. Noel Belcourt wrote:
>>>
>>> On Mar 24, 2010, at 1:18 PM, Rene Rivera wrote:
>>>
>>>> K. Noel Belcourt wrote:
>>>>>
>>>>> On Mar 23, 2010, at 3:59 PM, Rene Rivera wrote:
>>>>>
>>>>>> Vladimir Prus wrote:
>>>>>>> Rene Rivera wrote:
>>>>>>>
>>>>>>>> Belcourt, Kenneth wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> The log files for various trunk toolsets show this error.  Can
>>>>>>>>> someone
>>>>>>>>> please take a look?
>>>>>>>> The only immediate guess is the new bjam version. So I reverted
>>>>>>>> testing
>>>>>>>> back to 3.1.17 until I can figure out the problem.
>>>>>>>
>>>>>>>>> # Starting tests ("/scratch/boost/tools_bjam/bin.linuxx86_64/ 
>>>>>>>>> bjam"
>>>>>>>>> "-sBOOST_BUILD_PATH=/scratch/boost:/scratch/boost/tools_bb"
>>>>>>>>> "-sBOOST_ROOT=/scratch/boost/boost" "--boost=/scratch/boost/ 
>>>>>>>>> boost"
>>>>>>>>> "--boost-build=/scratch/boost/tools_bb" "--debug- 
>>>>>>>>> configuration"
>>>>>>>>> -l300
>>>>>>>>> intel-11.1 -d2 preserve-test-targets=off --dump-tests -j8
>>>>>>>>> "--build-dir=/scratch/boost/results"
>>>>>>>>>>> "/scratch/boost/results/bjam.log"
>>>>>>>>> 2>&1)...
>>>>>>>>> # Running Boost.Build tests
>>>>>>>>> # Using bjam binary in '/scratch/boost/tools_bjam/ 
>>>>>>>>> bin.linuxx86_64'
>>>>>>>>> # Getting test case results out of
>>>>>>>>> "/scratch/boost/results/bjam.log"...
>>>>>>>>> boost_root: /scratch/boost/boost
>>>>>>>>> locate_root: /scratch/boost/results
>>>>>>>>> *****Error - No "boost-test" lines encountered.
>>>>>>>>>     (Usually occurs when bjam was envoked without the --dump- 
>>>>>>>>> tests
>>>>>>>>> option
>>>>>>>>>      or bjam was envoked in the wrong directory)
>>>>>>>
>>>>>>> FWIW, running trunk version of bjam from status dir does  
>>>>>>> produce the
>>>>>>> 'boost-test(XXX)'
>>>>>>> output.
>>>>>>
>>>>>> So perhaps it is what Richard mentioned about subversion being  
>>>>>> stupid.
>>>>>> And we all need to remove the bjam dir.
>>>>>
>>>>> Hi,
>>>>>
>>>>> Here's how I run the tests, my nightly scripts are setup to  
>>>>> toast all
>>>>> directories before running the tests.
>>>>>
>>>>> rm -rf tools_bb tools_bjam tools_regression tools_regression_src
>>>>> rm -rf results
>>>>> rm -rf boost
>>>>>
>>>>> Next I setup some paths and copy in the necessary files and then  
>>>>> run the
>>>>> tests like this.
>>>>>
>>>>> python run.py --proxy="http://wwwproxy.sandia.gov:80"
>>>>> --runner="Sandia-gcc-4.2.4" --bjam-toolset=gcc --pjl-toolset=gcc
>>>>> --toolsets=gcc-4.2.4 --bjam-options="-j16 link=static" >&!
>>>>> gcc-4.2.4.boost.log
>>>>>
>>>>> So what bjam directory do I need to remove, sorry if I'm being  
>>>>> obtuse?
>>>>
>>>> It would be the tools_bjam dir you are already removing. So I'm  
>>>> at a
>>>> loss as to why the new bjam didn't work for you :-(
>>>
>>> I'm available today for a while if you have any suggestions or  
>>> want me
>>> to try some things to diagnose this.
>>>
>>> Please let me know.
>>
>> You could try running the unit tests for bjam in the machine that's
>> causing the problem. And of course manually running the tests, with  
>> bjam
>> -n --dump-tests, to see if it is/isn't doing the output with the new
>> version of bjam.
>
> Noel,
>
> did the above suggestion sched any light on the problem? Were issues  
> localized
> to some specific machines -- if so, which? Or it was a general messup?
> Does the problem reproduces when "run.py" is invoked from a terminal,
> as opposed to whatever cron is used to run them normally?

Sorry, I've had no chance to check this.  I'll get to this tomorrow  
and reply to the list then.

-- Noel