Re: [Boost-build] Doc build failure on release branch

Subject: Re: [Boost-build] Doc build failure on release branch
From: Vladimir Prus (ghost_at_[hidden])
Date: 2008-10-09 01:36:15

On Thursday 09 October 2008 08:49:25 Eric Niebler wrote:
> Vladimir Prus wrote:
> > On Thursday 09 October 2008 03:28:22 Eric Niebler wrote:
> >> Daniel James wrote:
> >>> I tried using my patch to build the
> >>> documentation, and ran into this:
> >>>
> >>> http://svn.boost.org/trac/boost/browser/trunk/tools/build/v2/tools/common.jam?rev=48570#L510
> >>>
> >> I have patched my copy of common.jam locally to work around this 
> >> problem. The diff is:
> >>
> >> Index: tools/common.jam
> >> ===================================================================
> >> --- tools/common.jam    (revision 49130)
> >> +++ tools/common.jam    (working copy)
> >> @@ -505,6 +505,10 @@
> >>       {
> >>           return "set $(variable)=$(value)$(nl)" ;
> >>       }
> >> +    else if [ os.name ] = CYGWIN
> >> +    {
> >> +        return "export $(variable)=$(value)$(nl)" ;
> >> +    }
> >>       else
> >>       {
> >>           # (todo)
> >>
> >>
> >> It seems to work just fine for me. I'm not sure if it's the Right Fix. 
> >> Perhaps someone more knowledgeable can have a look.
> > 
> > Can you explain what problem you originally run into, and what environment
> > you have. I think the comment was from Jurko, and it seems strange that this
> > new code works, while the old one does not, for any definition of "works".
> 
> 
> The environment this problem exposes itself is cygwin. 

Oh, well. I wish I understood why people keep using this weird beast :-(

> The problem is that environment variables cannot be set correctly. 

This is vague -- is the variable not set, or set to wrong value, and if so to what,
or error is emitted?

> It's a problem for  
> the doc build because before xsltproc is invoked, the CATALOG 
> environment variable needs to be set to point to a particular file; 
> otherwise, xsltproc cannot resolve certain stylesheet URLs correctly.
> 
> At least, that's my recollection, but Daniel has more recently grappled 
> with this problem. Maybe he can jump in.

In fact, the original

        return "$(variable)=$(value)$(nl)export $(variable)$(nl)" ;

probably is too complex, as

        return "export $(variable)=$(value)$(nl)" ;

should work everywhere, too. But anyway, I need to know the exact error that your,
or Daniel, have got building docs.

- Volodya