Re: [Boost-testing] Generating regression result tables

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2007-01-29 06:56:20

Vladimir Prus writes:
> First questions:
> 
> 1. I don't see either build_reports.py or script_runner.py in CVS, 

They are not (mostly for historical reasons).

> and I see some other scripts that are supposedly out-of-date. If
> not, can you add those two files to CVS?

Sure, but, strictly speaking, they are not concerned with and
therefore not required for producing report pages from local results:
'script_runner.py' is basically a generic logging script runner, and
'build_reports.py' in the end calls 'boost_wide_report.py', which is
the actual worker script (and it is in the CVS). You can invoke the
latter manually almost as easily:

    python boost_wide_report.py \
        --locate-root=$BOOST_ROOT \
        --tag=CVS-RC_1_34_0 \
        --expected-results=$BOOST_ROOT/libs/expected_results.xml \
        --failures-markup=$BOOST_ROOT/status/explicit-failures-markup.xml \
        --results-dir=<results directory> \
        --user="" \
        --dont-collect-logs

where

    <results directory>/incoming/CVS-RC_1_34_0/

contains the zip archive(s) of the XML results you want to produce the
reports for.

> 2. The email say to put a copy of Boost somewhere, but in practice,
> the script checks out the right parts from CVS. 

Oh, right.

> Is this expected, or putting a copy of Boost is no longer required?

It's not required for running the full-blown cycle through
'build_reports.py'. It's required for invoking 'boost_wide_report.py'
manually (obviously).

> 3. Is there a way to run those scripts on a local xml file, without
> putting things on FTP?

Not 'build_reports.py'; '--dont-collect-logs' does the trick for
'boost_wide_report.py'.

HTH,
-- 
Aleksey Gurtovoy
MetaCommunications Engineering