Re: [Boost-build] install a set of sources

From: Jurko Gospodnetić (jurko.gospodnetic_at_[hidden])
Date: 2008-07-19 04:22:18

   Hi Xavier.

> rule get_dependencies ( package ) {
>   ECHO "Boos dependencies from: " $(package) ;
>   SHELL "bcp --list --boost=/usr/local/include/boost-1_35 $(package)" ;
> }
> 
> local boost_files = [ get_dependencies scoped_ptr ] ;
> 
> local result ;
> for local b_file in $(boost_files) {
>   result += "-----> /usr/local/include/$(b_file)" ;
>   return $(result) ;
> }
> 
> install dist-src
>   :
> #  [ glob-tree *.cpp  *.h ] // This works correctly
>   $(result)
>   :
>   <install-source-root>.
>   <install-dependencies>on <install-type>H
>   ;
> ===============
> 
> The output is:
> 
> ===============
> Testing with:  scoped_ptr
> ----->  /usr/local/include/boost/assert.hpp
> boost/checked_delete.hpp
> boost/config.hpp
> boost/config/abi/borland_prefix.hpp
> ...
> ===============
> 
> So, seems that the "for" iteration only works correctly for the first
> element of $(boost_files). Why ?

   I just now took a look at this and it is not that the "for iteration" 
is processing only the first file on the list that is causing the 
problem but the fact that your SHELL invocation actually returned only a 
single item containing all the file names separated by new-lines.

   Therefore the "for iteration" processed that single item and simply 
added your -----> mark in front of it.

> And, is there any other way easier than this ?

   I have tried playing around with splitting the SHELL result into 
separate items but have been unsuccessful. I have been trying it on 
Windows that use the \r\n end-of-line mark instead of \n so that might 
have been the cause but I do not believe this can be done portably with 
the current Boost Jam.

   Its MATCH rule & therefore its regex.split rule does not seem to be 
able to split text containing newlines into separate items - one per 
line. It could be that could be made to work by adding support to Boost 
Jam to directly represent end-of-line characters in code, but did not 
take the time to try this out.

   I guess it could be done in an unportable way by using a construct like:
     local nl = "
" ;
     local match = [ MATCH "^([^$(nl)]*)[$(nl)]?(.*)" : $(string) ] ;
   But you would have to play around with that to see something like 
that helps. Note just that this would consider \r\n to mean two 
end-of-line characters.

   Sorry if this was not helpful enough. Perhaps someone else can be of 
more help and perhaps someone can test this for you on some *nix OS.

   Best regards,
     Jurko Gospodnetić