|
|
Boost Testing : |
From: Rene Rivera (grafik.list_at_[hidden])
Date: 2005-07-16 00:59:47
Robert Ramey wrote:
> Rene found the magic to make the majority of the serialization test pass.
> However, the tests which use "export" still don't pass. I still believe
> there is overzealous optimization by the linker? which eliminates code for
> types not explicitly referred to by name. Using he mw -help comand I see
> there is a swithc [no]deadcode. This would seem to be relevant to the
> situation here. Its not clear what the default is and what the switch
> setting should be or even if its relevant.
That is an optimization control option. It refers to standard local dead
code optimizations. Even if it was relevant, since we (usually) test
with the debug variant, all optimizations are off anyway.
I know where it's not generating the code you are expecting it to
generate. But I have no real clue how to fix it. I'll mention more
tomorrow when I have a bit more time to track it down again (it takes a
fair amount of time to single step through all that serialization code).
-- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - Grafik/jabber.org