RE: [Boost-testing] [Wave] Regressions, need some help

From: Hartmut Kaiser (HartmutKaiser_at_[hidden])
Date: 2005-04-29 11:36:38

Rene Rivera WROTE.

> Hartmut Kaiser wrote:
> > 1. RSI-SoL gcc-3_4_3-linux-i686-release (Rene?)
> 
> Yes :-)

:-) Simply wasn't sure...

> > The exit code of the Wave test application is reported to 
> be 'EXIT STATUS:
> > 137', which seems to be a result of a sig 9 (128 + 9). It's 
> difficult 
> > for me to reproduce this. Would it be possible to get a stack 
> > backtrace to narrow down the point of failure?
> 
> It would be possible.. But since it's running a release build 
> it's not going to be much help as it doesn't have symbols. 
> But I'll see what I can do; perhaps I can independently run 
> it in profile mode and see if the error still exists.

How do you achieve the release mode builds? Overwriting the flags on the
command line? I'm asking, because at least the Wave-test Jamfile explicitely
specifies to build the debug modules.
But this gives me a clue. I'll try to compile it here with my gcc in release
mode (optimisation on) but with debug symbols, let's see what happens ;-)

> > 2. LaruCorp vc-8_0 (Carl?)
> > The logs report a missing 'libboost_filesystem-vc80-mt-s-1_32.lib', 
> > which should be generated by bjam because of the specified 
> > dependencies (and which _is_ generated for other toolsets). 
> I honestly 
> > don't understand, why this library may be missing... Any clues?
> 
> Ah.. Which logs? For which test?
> 
> Oh never mind.. Just noticed you are talking only about the 
> wave tests :-)

Yep. I'm currently trying to figure out the Wave problems on different
platforms.

> > 3. Dirxion cw-9_3-darwin and gcc-3_3-darwin (Stephen?) Both 
> toolsets 
> > report missing (though different) external symbols which is 
> behind my 
> > understanding as well, since all of the symbols reported as missing 
> > should be available through linked system or boost libraries. Any 
> > ideas?
> 
> I've been looking into the cw-9_3-darwin in the past few 
> days. Many of the missing symbols are of the ctype<> variety. 
> But I've been unable to figure out why they are missing. When 
> I build the tests from the IDE it has no problems finding the 
> symbols. But when using the command line tools it doesn't 
> work :-( I was down to doing diffs of preprocessed code to 
> see what I could decipher.

Yeah, the cw-9_3-darwin related missing symbols are from the ctype<> family.

> And of course scratch the above, now that I noticed that you 
> are talking about the wave tests only. I'll try and see if 
> it's a similar problem as the ctype problems.

Thanks a lot!
Regards Hartmut