Re: [Boost-build] Understanding generated files / targets.

Subject: Re: [Boost-build] Understanding generated files / targets.
From: Tom (tabsoftwareconsulting_at_[hidden])
Date: 2014-07-28 00:18:07

Steven Watanabe <watanabesj <at> gmail.com> writes:

> On 07/27/2014 12:42 PM, Tom wrote:
> > I am trying to generate a C++ source (or header) file containing
> > versioning information generated from 'git describe'.  This has become
> > necessary as we have moved to git and a branched development model and
> > modifying a version file manually is a non-starter (due to conflicts
> > on merg) in this development model.  We will be driving the version
> > number in binaries by tags in revision control.
> > 
> > This is challenging mainly in that the generated file depends on state
> > not known (or knowable) by Boost.Build as opposed to a normal file
> > transformation.  This results in either not generating the file when
> > it should be (a tag is checked out or a file is edited which do not
> > affect the code handling the version number, then the build is run) or
> > in rebuilding things when it is not actually necessary (which causes
> > all tests to be re-run for example).
> 
> If you can use the print module to generate the file,
> then it automatically checks whether the content has
> changed.  (See b2 --help print for information on
> how to use it)

Steven,

That worked really well, I wish I knew that little trick before :)

It took a little while for me to get it right, so I've included a minimal
Jamroot that shows how it was done for others that might need
something like this.

Thanks!
Tom

# Jamroot
import print ;

explicit generated ;
exe generated : generated.cpp ;
explicit generated.cpp ;
make generated.cpp : : @generate-file ;
rule generate-file ( target : sources * : properties * )
{
   print.output $(target) ;
   print.text "int main() { return 0; }
" : true ;
}