Need help understading boost regression test

[Mario F.]

I'm getting a warning when compiling with boost::filesystem under mingw-g++ 3.4.5 in debug mode.

The warning is:

passing negative value '-0x000000001' for converting 3 of 'boost::filesystem::basic_filesystem_error<Path>:: basic_filesystem_error(const std::string&, const Path&, boost::filesystem::system_error_type) [with Path = boost::filesystem:ath]'

I took a look in the web and was sent to the following page:

Regression test

However, I can't read it. I'm not familiar with the concept of regression tests what they mean exactly and what can be done with - or learned from? - the information.

What I need to know is to first understand the compile time warning.

It points to a negative value where it seems it expects a positive one. However I can't seem to dig in enough the boost code to access the boost::filesystem::system_error_type declaration. So i'm left guessing there is an intended conversion taking place but mingw flags it regardless.

I would also need help in understanding what exactly is a regression test from the boost perspective.

What troubles me here is the fact this warning is flagged by the regression test as success. So my understanding of regression tests as described on the wiki may not be enough.

[CornedBee]

The regression tests are a set of small programs that test various invariants in the libraries. Basically, each test tests a small part of the functionality of the libraries.

For each test, it is defined what the expected result is. Some tests are expected to fail compilation. Some tests are expected to succeed in compilation but generate some error at runtime. Some tests are expected to run cleanly.
Boost does not, at this time, ensure that no compiler warnings are created during compilation of the tests. There are various reasons for this, but the primary one is lack of resources and time.

Nevertheless, passing a -1 compile time constant to unsigned types (system_error_type is an unsigned integral type) is commonly accepted practice to get the largest possible value of that type, so this warning is probably spurious.

[Mario F.]

I completely forgot about this. Sorry, CornedBee. ++Extremely busy these past months. But still it's not my way to let an answer go without at least a thank you.

Why you think numeric_limits wasn't used instead?

[CornedBee]

*shrug*
Because it's shorter, probably.

[Mario F.]

Yes... anyways, thanks.