|
|
Boost Testing : |
From: AlisdairM (alisdair.meredith_at_[hidden])
Date: 2006-11-16 08:47:07
Anthony Williams wrote:
> Secondly, I would expect most applications to be compiled with what
> amounts to "release" options --- full optimization, no debug symbols,
> etc. If boost code doesn't work under these circumstances, that's
> problematic.
If the compiler does not work under those situations it is even more
problematic! The Borland environment can be 'challenging' until you
adapt to its idiosyncracies.
One of those is to only ever build staticly linked, full-debug
releases. Anything else is prone to hard-to-diagnose problems. Those
problems can be found and nailed one by one, but it is rarely worth the
effort - the performance and image size is usually not that different.
For instance, Borland don't embed their debug info in the .exe, but
generate a companion resource file (tds extension) The debug mode
might not be as conservative as other environments - certainly my
Borland debug builds run significantly faster than MS debug builds,
just as MS release builds run significantly faster than Borland release
builds.
I agree release testing is important, on all platforms, but I am not
too hung up on Borland release mode failures at this point.
Hopefully their next iteration of compilers may make this a reasonable
option.
[Although the entire Borland tools division just got spun out into a
new company called CodeGear, so we may need to rename this compiler
soon as well - hopefully not until after 1.34 is released though!]
-- AlisdairM