From: Howard Hinnant (hinnant_at_[hidden])
Date: 2004-08-26 08:47:15
On Aug 26, 2004, at 7:28 AM, David Abrahams wrote:
> "Gennadiy Rozental" <gennadiy.rozental_at_[hidden]> writes:
>>> all program_options tests fail on cw 8.3, because boost.test refers
>>> to the
>>> _errno variable, but that reference is somehow unresolved.
>> I don't think I refer to this veriable (at least directly). Something
>> compiler related.
> It appears to be related to the use of std::setw. Maybe you could
> use some hacks to try to force it to recognize errno, e.g. directly
> referencing it? Howard, can you think of any way to approach
> debugging this?
I would start with a HelloWorld. If that works I would gradually move
stuff from the test into the HelloWorld until I could replicate the
problem. That process might identify what is causing the problem.
std::setw is a pretty simple inlined mechanism. It does nothing but
set a integral variable in the ios_base. This may be a red herring.