From: Jurko GospodnetiÄ (jurko.gospodnetic_at_[hidden])
Date: 2008-07-22 18:29:34
>> b) Toolset initialization is supposed to not cause bjam to exit even
>> if it fails, right?
> Yes, it is stated at
Hmmm, I never thought about this much but never found it to be wrong
that an incorrect tool configuration causes the build to fail.
For instance, why should not a tool configuration fail in case it is
called multiple times with conflicting configuration options? In that
case the configuration can never be correct under any circumstances.
One such example would be configuring the same version of a compiler
toolset each time with a different executable.
As initializing a toolset is a site level configuration option, I do
not see how the rationale given for this rule on the cited web page
applies: 'Configuration may be shared between different machines, and
wrong values on one machine can be OK on another.'.
We do not initialize toolsets on different machines using the same
file. And even if we were to use such a 'shared' configuration file it
would be cleaner to add code to it that detects which machine it is on
and adjust the initialization accordingly.
I'd be most interested in hearing other people's opinions on this.
Just my 2c worth...
I'll look into correcting this ugly msvc toolset error message. It
really should not fail with such an 'internal error' as not passing the
correct number of parameters to some rule call.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk