Re: [boost] [integer] Can anyone double-check this test? (FIXED?)

From: Daryle Walker (darylew_at_[hidden])
Date: 2008-07-07 04:54:39

[I tried to send this out an hour ago, but I sent it through the  
wrong e-mail address.  That ended up being a good thing since I fixed  
the problem!]

On Jul 7, 2008, at 12:15 AM, Vladimir Prus wrote:

> Daryle Walker wrote:
>
>> On Jul 6, 2008, at 7:35 AM, Vladimir Prus wrote:
>>
>>> Daryle Walker wrote:
>>>
>>>> OK, I tried Boost.Build out:
>>>> [SNIP]
>>>> 2. I put $MY_BOOST down as my $BOOST_BUILD_PATH, so I wouldn't have
>>>> to put the *-config.jam files in my home directory.  My copies  
>>>> of the
>>>> latest Boost download and the subversion repository are also in  
>>>> this
>>>> directory.  Will that mess anything up?  What's the "site- 
>>>> config.jam"
>>>> file for anyway?  Only one line was uncommented in "user- 
>>>> config.jam",
>>>> the "using gcc ;" line.  Is this suitable for a Mac OS X (10.4.11)
>>>> system?
>>>
>>> No, you need "darwin" as tools.
>>
>> Do you mean changing the "using gcc ;" line to "using gcc : : :
>> <linker-type>darwin ;"?  But doing "bjam-svn --help gcc.init" makes
>> it seems that the "linker-type" should have been chosen by default?
>
> No, I mean "using darwin ; "
>

Rene also sent a response to my last e-mail, I've changed my file to  
add a "using darwin ;" line.  Since reading this letter,...

>> Hmm, doing that change made it work!  However, I got a help message
>> from doing "bjam-svn --help darwin.init" too.  Does this mean I
>> should add a "using darwin ;" line in my user-config.jam?  Would that
>> line complement or replace the current GCC line?
>
> The version of gcc provided by apple goes by the name 'darwnin'. You'd
> need 'gcc' toolset only if you have a FSF gcc somewhere.

..I've commented out the GCC line.

>>>> daryle[hello]$ bjam-svn toolset=darwin
>>>> ~/Documents/Programming/Boost/main-svn/boost-trunk/tools/build/v2/
>>>> build/feature.jam:466: in feature.validate-value-string from module
>>>> feature
>>>> error: "10.2" is not a known value of feature <macosx-version>
>>>
>>> Oh, you actually have 10.2 SDK installed? Can you "svn up" and try
>>> again?
>>
>> Building the Hello example after my last SVN update didn't change
>> anything.
>
> Is the error message *exactly* the same? Or it now complains about  
> 10.1?

I don't think it did, but I had to reset my computer between runs.

>> Are you assuming that I'm actively using the 10.2 SDK?
>> I'm not, but...
>>
>> ..I have all of the SDKs I can support.  I'm on a Mac OS X 10.4
>> (PowerPC) system.  The versions of XCode that my OS supports can
>> retain multiple SDKs.  Besides the native 10.4 (Universal) SDK, I
>> also have the SDKs for 10.3.9, 10.2.8, and the super secret 10.1(.5?)
>
> Oh, you have 10.1 too? Please update, make sure tools/build/v2/ 
> tools/darwin.jam
> is of revision 47164 and try again.

OK, I did another update and have revision 47170.  Going to the  
Boost.Integer test directory and running bjam....  Nope, the same  
failure results that I posted in response to Rene, except that the  
number of updat(ing|ed) targets has reduced by one to 42 & 24,  
respectively.  Hmm, I added the "using mpi ;" line to my user- 
config.jam, but I still get warnings about it.  That and the fact  
that the warnings also complain about that there's no toolset  
configured makes me think that my configuration file is being ignored.

Let's see what happens when I retry my (seemingly) successful Hello  
example again:

//=====================================
daryle[hello]$ ls
Jamroot         hello.cpp
daryle[hello]$ bjam-svn
MPI auto-detection failed: unknown wrapper compiler mpic++
Please report this error to the Boost mailing list: http://www.boost.org
You will need to manually configure MPI support.
..found 9 targets...
..updating 5 targets...
MkDir1 bin
MkDir1 bin/darwin-4.0.1
MkDir1 bin/darwin-4.0.1/debug
darwin.compile.c++ bin/darwin-4.0.1/debug/hello.o
darwin.link bin/darwin-4.0.1/debug/hello
..updated 5 targets...
daryle[hello]$
//=====================================

Here, it paid attention to the MPI line I put in the file.  (Of  
course, since it doesn't work, that means that I have a new problem  
now....)  And the "hello" program works!  Why isn't the regular case  
reading my configuration file?

POST-MAIL-SCREW-UP ADDENDUM:

I got it to work.

The warnings about no configuration triggered something.  I recalled  
that I put my user-config.jam file in my general Boost directory and  
used the $BOOST_BUILD_PATH shell variable to let bjam know where the  
file is.  It is an alternative to putting the configuration file in  
my home directory.  (I used to have it there in a previous attempt.)   
As soon as I put a soft-link to user-config.jam in my home directory,  
the Boost.Integer tests worked!

This isn't good, though.  The BOOST_BUILD_PATH method worked just  
fine for the Hello example.  This means that the Boost.Integer test  
Jamfile, or one of the Jam-files above it in the hierarchy, broke the  
BOOST_BUILD_PATH capability.  Can you check the Boost.Integer test  
file to make sure that's not the problem?  If it is, let me know how  
to fix it.  Otherwise, I'll leave the hierarchical file problem to  
you.  (I made a copy of this problem as ticket #2082.)

-- 
Daryle Walker
Mac, Internet, and Video Game Junkie
darylew AT mac DOT com