Re: Win32 regression test failures

From: Daryle Walker (darylew_at_[hidden])
Date: 2001-11-30 18:54:13

on 11/30/01 3:01 PM, Peter Dimov at pdimov_at_[hidden] wrote:

> From: "Beman Dawes" <bdawes_at_[hidden]>
>> A fresh copy of the Win32 regression tests have just been uploaded to
>> http://boost.sourceforge.net/regression-logs/win32.zip
>> 
>> The attached abbreviated table is for only tests which changed (compared to
>> 1.25.1).
>> 
>> There are new passes in config_test and random_test.
>> 
>> There are new failures in bind_test, mem_fn_test, and function_test.
> 
> Comeau and Intel fail because void returns don't work, but
> BOOST_NO_VOID_RETURNS is not defined (they fail config_test as well.)
> 
> Both compilers can be configured to allow void returns, I believe - probably
> their MS compatibility mode turns void returns off.
> 
> The Metrowerks failure has been #ifdef-ed out in the latest version of
> bind_test.cpp.
> 
> The MSVC failures are due to the internal compiler error with debug info on.
> They, too, are addressed (by turning off void return support in mem_fn.hpp)
> by the latest versions in the CVS.
> 
> Please update and re-test.

Metrowerks has special preprocessor macros (like "__option(XXX)") to let you
know the state of a particular option.  If Comeau and Intel can sometimes
support void returns and sometimes not, hopefully they have similar
preprocessor macros to detect their option states.  If so, you can define or
not define BOOST_NO_VOID_RETURNS based on that option.

-- 
Daryle Walker
Mac, Internet, and Video Game Junkie
darylew AT mac DOT com