Re: [boost] [Concepts] stl_concept_covering test

From: David Abrahams (dave_at_[hidden])
Date: 2005-09-13 12:49:36

Doug Gregor <dgregor_at_[hidden]> writes:

> This test is failing on a whole lot of compilers. For instance, see:
>
> 	http://engineering.meta-comm.com/boost-regression/CVS-RC_1_33_0/ 
> developer/issues.html
>
> Actually, a quick review shows that it fails on every compiler out  
> there :) Normally, I'd blame the test itself, but given that this test  
> was meticulously crafted to poke at the corner cases of STL  
> implementations, it's very likely that the test is correct. Would  
> someone volunteer to check out the errors and mark failures in  
> explicit-failures-markup.xml appropriately?

Last week as Matthias Troyer and I were preparing slides for a class,
we decided to try out one of the examples from the test.  It turned
out that the test wasn't in the test suite.  

Here's an excerpt of a chat with Jeremy:

  boostdave:	Hi, J!

  jeremysiek:	Hi Dave!

  jeremysiek:	Are you in Switzerland?

  boostdave: You bet!  Matthias just found a problem with the concept
             check tests

  jeremysiek:	cool (about being in Switzerland, not the problem!)

  boostdave: Yes, very.  --- The stl_coverage test is invoked in
             libs/concept_check/Jamfile

  boostdave: However, status/Jamfile just goes on running the other
             concept_check tests and never subincludes the Jamfile

  boostdave:	The coverage test is failing for matthias when he tries to run it.

  boostdave: Do you want me to fix status/Jamfile so you start getting
             error reports?  ;-)

  jeremysiek:	sure

  jeremysiek:	so yeah, thanks for offering to fix the Jamfile

  boostdave: OK.  It's going on the release branch too, since after all
             this is the time to fix bugs for the point release.

  jeremysiek:	nod

  jeremysiek: just one thing to note, stl_concept_covering.cpp is more
              of a test of the C++ implementation than of the concept
              checking lib

  jeremysiek:	so often failures are expected

  boostdave:	Unless you got a bug in your archetype.

  jeremysiek:	true

  jeremysiek:	or a bug in the test itself

  boostdave: But anyway, you can just mark the explicit failures markup
             XML

  jeremysiek:	ok

  boostdave:	and then report the bug to GCC ;-)

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com