Boost Interest :
Subject: Re: [Boost-cmake] CMake for development testing
From: Doug Gregor (doug.gregor_at_[hidden])
Date: 2009-06-09 17:49:26
On Mon, Jun 8, 2009 at 7:30 AM, Bill Hoffman<bill.hoffman_at_[hidden]> wrote:
> Brad King wrote:
>> David Abrahams wrote:
>>> 67/ 79 Testing Python-result Passed
>>> 68/ 79 Testing Python-string_literal ***Failed
>>> 69/ 79 Testing Python-borrowed ***Failed
>>> 70/ 79 Testing Python-object_manager Passed
>>> 71/ 79 Testing Python-copy_ctor_mutates_rhs Passed
>>> 72/ 79 Testing Python-upcast Passed
>>> Worse, it's in there among all the successful test output, so I have to
>>> dig around to find the failures.
>> The list of failures is summarized at the end of the output, isn't it?
>> Then you can use '-R' to regex-select tests and '-V' to see verbose
>> ctest -R Python-string_literal -V
>> We've found this basic command-line approach to be sufficient because:
>> 1.) The display on CDash from "make Experimental" is interactive.
>> We typically submit to the main server, but it is possible to
>> submit to localhost too.
>> 2.) When fixing a test we often need to run it specifically many
>> times, so typing a command-line to select one test is necessary.
>> For these reasons we've never felt the need for better reporting from
>> the command-line tool. If you have a suggestion to improve reporting
>> from a one-off command-line run, please post it. However, we don't
>> want to get into advanced html generation which duplicates work already
>> done in CDash.
> In CVS CMake there is also a new feature, if you set the environment
> variable CTEST_OUTPUT_ON_FAILURE, it should show the full output of those
> tests that fail.
Ah, that's perfect for my use cases. Thanks!