Boost logo

Boost-Build :

From: Kirill Lapshin (klapshin_at_[hidden])
Date: 2003-11-07 00:23:34


Hi Volodya,

> > warning: using independent target
> > <..!__libname__/msvc-7.0/debug>__libname__.rsp
>
> That's an interesting thing. It means that bjam is creating a target
that
> nobody depends on. Right, nobody depends on rsp. Anybody knows why the
> warning is there? I'm all for removing it completely.

I've tried m9.1 and had to rollback to m8. It is a very weird problem
I am seeing here. Basically I have a project which depends on boost
libraries. One (static) library uses internally (completely hidden
within cpp file) a map from string to boost::shared_ptr. One of the
functions return object stored within shared_ptr -- it simply looks up
map and dereferences a smart_ptr. This library linked to test
executable. When test runs it deadlocks in smart_ptr destructor in
lightweight mutex. The weird part is that if you compile it with m8
everything works fine. If compiled with m9.1 it hungs, if compiled
with m9.1 but without a static library (directly compiling cpp into
exe), it works fine again. It hungs only if compiled with msvc, with
gcc it works fine.

I compared all compile/link flags, and they are almost identical. All
the difference I see betwenn 8 and 9 is /INCREMENTAL:NO in m9.1, and
fewer -I options -- directories where obj genereated are not included
anymore. I am totally lost. I tried to create a simple testcase to no
avail. The only thing which looks weird in m9.1 is this rsp warning,
and looks like my problem is triggered by link somehow -- because
compiling cpp directly into exe works fine, while through lib fails --
in both cases cpp compiled with exactly same flags (well it is
compiled into different folder so some paths are different). So I was
wondering if rsp warning has something to do with the problem. Why we
have not seen it before?

I understand that it is almost impossible to diagnose the problem
using only this vague description, but if anyone have
ideas/suggestions how to tackle it, I would be very grateful.

--Kirill

 


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