From: Daniel Lord (daniel_at_[hidden])
Date: 2008-03-01 17:36:31
UPDATE:Just came across this in the Boost documentation for filesystem which
might be an explanation:
"All exceptions thrown by the Filesystem Library are implemented by calling
boost::throw_exception()<file:///Users/daniello/Documents/Development/Lib/Boost/boost_1_34_1/libs/utility/throw_exception.html>.
Thus exact behavior may differ depending on BOOST_NO_EXCEPTIONS at the time
the filesystem source files are compiled."
Daniel
---
Some people building C++/Obj-C code on OS X with Xcode and have a problem
think: "I know, I'll use the Boost Libraries". Now they have two
problems--three if they used Boost for regular expressions. ;-)
On 3/1/08, Daniel Lord <daniel_at_[hidden]> wrote:
>
> Lothar,
> My terse reply was not meant to be disrespectful ( in case you took it
> that way--text conversations can be misunderstood). I only posted the output
> in case you already figured it out thus avoiding spamming the list with a
> huge post given there don't seem many OS X developers on it so very few give
> a fig about this issue. But our numbers are growing and these painful
> learning experiences are becoming more common. The regime below is a
> quick-and-dirty sub-optimal brute force way of achieving a universal fat32
> C++ build on OS X with Xcode 3. Once working, you can refine from there. I
> am no expert nor journeyman with Xcode though I hope to be one day. This
> works for me so I can focus on my application work. HTH.
>
> ARCHITECTURE ISSUES
> ----
> Both my systems run Leopard 10.5.2 and are i386 ( a Core Duo Macbook Pro
> and a first-gen Quad core 8-way SMP Mac Pro).
>
> My Boost build seem to produce fat32 binaries "out of the box" with no
> special configuration. as the out of 'file' on the boost_filesystem libs
> shows:
>
> ./static/debug/mt/libboost_filesystem-mt-d-1_34_1.a: current ar archive
> ./static/debug/mt/libboost_filesystem-mt-d.a: current ar archive
> ./static/debug/st/libboost_filesystem-d-1_34_1.a: current ar archive
> ./static/debug/st/libboost_filesystem-d.a: current ar archive
>
> ./static/debug/st/libboost_filesystem-gcc40-d-1_34.a: Mach-O universal
> binary with 2 architectures
> ./static/debug/st/libboost_filesystem-gcc40-d-1_34.a (for architecture
> ppc): current ar archive
> ./static/debug/st/libboost_filesystem-gcc40-d-1_34.a (for architecture
> i386): current ar archive
>
> ./static/release/mt/libboost_filesystem-gcc40-mt-1_34.a: Mach-O universal
> binary with 2 architectures
> ./static/release/mt/libboost_filesystem-gcc40-mt-1_34.a (for architecture
> ppc): current ar archive
> ./static/release/mt/libboost_filesystem-gcc40-mt-1_34.a (for architecture
> i386): current ar archive
>
>
> IMPORTANT: notice not all the libraries are fat32 architecture--make sure
> you link in fat32 binaries (the one that the 'file' commands states are
> "universal binary with 2 architectures) if you want a universal (n this case
> fat32) build. You will also need to adjust your target build settings to
> include the architectures you want to build for. read teh Xocde docs to
> learn how to do this. I have tried tweaking Boost to give me a fat quad
> build (i386, ppc, ppc64, x86_64) yet because of the use of bjam.
>
>
> If you are not having this occur, it might be that you have CFLAGS<
> CXXFLAGS, and LDFLAGS set to "-arch i386" rather than "-arch i386 -arch ppc"
> somewhere in your environment. It might help to set them explicitly in the
> shell and export them before running Boost though Boost might override them.
> Bjam is too arcane for me to try and figure out how Boost build works--I
> just don't have the time avaialbe with all the other things I need to learn
> more about and projects I need to work on.
>
> PATHS
> -----
> At first I tried using the default install location for the Boost
> libraries (/usr/local) but that was problematic with Xcode since it seemed
> to grab the first libboost_filesystem binary is found and it would be of the
> wrong architecture. Longer term, I am hoping that delving into the Xcode
> documentation will produce a more elegant solution to this, but for now I
> jsut use brute force and copy the files I need to the project folder. If
> you use source control this has the advantage of freezing the dependent
> library versions for SCM archiving--few things more vexing than having a
> library change versions and break your code. Specifically, I copy the
> relevant files to the project root directory, the path in the Xcode
> environment variable $(SRCROOT):
>
> 1)I copy all the boost includes into my project root folder. Copy the
> 'boost' folder inside boost_1_34_1, once you get it working you can pull out
> the ones you dont need. Also you could probably put the m in a silgne
> diretory hierarchy rooted somewhere below your home to avoid the copying,
> but you can work that out later--I want to keep things simple now to get it
> working first.
>
> 2) I copy the particular libraries I need in a folder named "lib" in
> $(SRCROOT). See above in ARCHITECURES to make sure you copy the right files.
> Check them iwth the 'file' command to be sure they are universal binaries of
> the desired type.
>
> 3) Modify your build settings for the target (right-click on the target in
> the Xcode project view and select the build tab). lot of settings there to
> tweak, but you jsut need to setHEader Search Paths and Library Search Paths
> to $(SRCROOT). IMPORTNT: do not check recursive or it whacks your name sapce
> and you'll get error on building like " 'snprintf' is nto a memeber of
> 'std'". this dorve me nearly insane when it first happened until I figured
> it out. I should have RTFM'd and could have avoided the pain and anguish.
>
> 4) open the target tree to expose the build phases (click on the trainagle
> next to the target) and you'll see three standard phases: Compile Sources,
> Link Binary With Libraries, and Copy Files. right-click on Link Binary With
> Libraries and selet add existing file and select you libraries from the file
> choose dialog.
>
> If this is a all done correctly you should get a fat32 copile fo your
> application that works. At least it works for me. As I have said repeatedly,
> this is a quick and dirty regime that works but is sub-optimal. Over time
> I'll refine it to ake it more efficient and maintainable. But to goal here
> is to get you moving with Boost and Xcode universal builds.
>
> I am no expert but this worked for me.
> HTH
>
> Daniel Lord
> 1 March 2008
>
> ---
> Some people building C++/Obj-C code on OS X with Xcode and have a problem
> think:"I know, I'll use the Boost Libraries". Now they have two
> problems--three if they needed regular expressions. ;-)
>
>
> On 3/1/08, Lothar May <boost_at_[hidden]> wrote:
> > Hi Daniel,
> >
> > thank you very much for testing it. It's great that it works for you.
> > Could you please tell me how you built boost? It seems you are using
> the
> > i386 architecture. Is your boost build a universal build, i.e. will the
> > programs run on PowerPC as well?
> >
> > The problem for me is that I do not have a PowerPC based system, so I
> > cannot build boost on a PowerPC and on an i386 system and copy them to
> a
> > universal build - instead I need to compile both on the i386 system.
> The
> > way I tried to do this, although as documented, seems to break the
> > exception support. Maybe we can find a way to merge the way you are
> > building boost with the additional options needed for a universal
> build.
> >
> > Thanks again,
> >
> > Lothar
> >
> >
> > Daniel Lord wrote:
> > > It worked fine for me:
> > >
> > > [09:46:50] daniello_at_Mercury
> > > ~/Documents/Development/Study/Xcode-Boost/BoostExceptionTest
> > > $uname -a
> > > Darwin dhcp-2op9-2op10-east-130-35-99-215.usdhcp.oraclecorp.com
> > > <http://dhcp-2op9-2op10-east-130-35-99-215.usdhcp.oraclecorp.com>
> 9.2.0
> > > Darwin Kernel Version 9.2.0: Tue Feb 5 16:13:22 PST 2008;
> > > root:xnu-1228.3.13~1/RELEASE_I386 i386
> > > [09:46:57] daniello_at_Mercury
> > > ~/Documents/Development/Study/Xcode-Boost/BoostExceptionTest
> > > $./BoostExceptionTest /fake
> > > This is never printed on MacOS 10.5.
> > >
> > >
> > > On 2/26/08, *Lothar May* <boost_at_[hidden]
> > > <mailto:boost_at_[hidden]>> wrote:
> > >
> > > Hi Daniel,
> > >
> > > compiling boost is successful on my system, too. Are boost
> exceptions
> > > caught using your build? Is your build universal?
> > >
> > > > I missed the thread on this, but I'll help you out if I
> can--just
> > > > bring me up to speed.
> > >
> > > Sample test code is this:
> > >
> > > --cut here--
> > >
> > > #include <iostream>
> > > #include <boost/filesystem.hpp>
> > >
> > > using namespace std;
> > > using namespace boost::filesystem;
> > >
> > > int
> > > main()
> > > {
> > > try
> > > {
> > > directory_iterator i("does_not_exist");
> > > } catch(...)
> > > {
> > > cout << "This is never printed on MacOS 10.5." << endl;
> > > }
> > > return 0;
> > > }
> > >
> > > --cut here--
> > >
> > > Best regards,
> > >
> > > Lothar
> > >
> > >
> > > _______________________________________________
> > > Boost-users mailing list
> > > Boost-users_at_[hidden] <mailto:Boost-users_at_[hidden]>
> > > http://listarchives.boost.org/mailman/listinfo.cgi/boost-users
> > >
> > >
> > >
> > >
> > >
> > >
> ------------------------------------------------------------------------
> > >
> > > _______________________________________________
> > > Boost-users mailing list
> > > Boost-users_at_[hidden]
> > > http://listarchives.boost.org/mailman/listinfo.cgi/boost-users
> >
> > _______________________________________________
> > Boost-users mailing list
> > Boost-users_at_[hidden]
> > http://listarchives.boost.org/mailman/listinfo.cgi/boost-users
> >
>
>
> --
>
>
>
>
>