From: Stefan Slapeta (stefan_at_[hidden])
Date: 2004-02-18 09:01:54
Russell Hind wrote:
> Well it isn't boosts' fault. It is a limitation in the way win32
> handles thread cleanup and stuff and currently there isn't a work-around
> for it. I suggest bug reports to Microsoft rather than having a go at
> the volunteer boost developers.
It wasn't my intention to blame boost developers for a win32 defect -
sorry if anybody got this impression! It just regret seeing the current
boost library breaking all my thread projects; that's the point.
> Ok, this I agree with. There was no mention of this in the changes for
> 1.30.x and there probably should have been, but it is fairly obvious the
> first time you build boost::thread from 1.30.x that it isn't there, so
> stick with 1.29.0 until you are at a stage where distributing shared
> runtimes might be possible.
Yes and no!
As I mentioned, we talk about TWO changes here:
The first - disabling the static library for boost.thread - was made in
1.30.0. BUT: there was a (more or less dirty) workaround in disabling a
macro in thread/detail/config.hpp + building an own static library
without the tss class (and cpp file). This worked if you didn't use tss
(which many of us don't).
The second - forcing boost to use the shared rt lib - is new with
1.31.0. A possible workaround for this could be disabling the auto_link
macro in config/user.hpp.
It seems that now after applying these two 'patches' boost.thread
'works' again as static lib and, more important than that, without the
I consider the latter point very critical for large companies as it is
simply very expensive in administration to install the shared rt on
thousands of machines, each having its own os and configuration! Once I
have to do this also for the rest of the thread library (means: not just
for tss), this could be a clear economic argument against using
boost.threads in the future at all, which I would regret even more.
What I want to say: PLEASE enable the use of the static runtime in
boost.thread for users who don't need any other functionality than the
'pure' thread library, again!
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk