Re: [boost] Regression testing (1.33)

From: Jeff Garland (jeff_at_[hidden])
Date: 2005-05-31 21:00:26

On Tue, 31 May 2005 11:38:09 -0700, Victor A. Wagner Jr. wrote
> Since we seem to be getting close (date anyone?) to the release of 
> 1.33, I've changed my testing cycle to 8 times a day to improve 
> turnaround for the library authors. 

Thx -- that's very helpful -- hope the backend processing can keep up!

> The only remaining failures in 
> VC-8_0 are some datetime things (the localtime test is clearly a 
> deliberate choice on someone's part to make dates before 1970 
> "illegal").  Are there any representatives from either Microsoft or 

These are just warnings, you'll note that the test compiles and runs -- until
it blows up like the others...

> Dinkumware here who can explain this decision?  Should we mark them 
> as "expected failures" (though I'd object to having the block turn 
> green simply because _I_ beileve it is an error and should show up 
> as such)

I investigated this.  What they've done is deprecated the standard interface
because of 'security issues' and invented a new one that is 'safer'.  However,
there may be some cost in terms of speed.  In any case, I was reluctant to
move away from the standard interface just for this compiler at the last
minute before a release.  BTW, I'm pretty sure this happened in the last beta
upgrade b/c I don't remember seeing this b/f.

> the two "input facet" test failures show only:
> Run output [2005-05-31 17:13:49 UTC]:
> EXIT STATUS: 666
> 
> not a lot of information for chasing down the actual problem.

Yeah -- I put some debugging code in but to no avail. Any chance you could
trap it in the debugger and send me a stack trace?    I didn't pay much
attention earlier since when Carl was running with VC8 for awhile this was all
passing on his machine...

Jeff