From: Douglas Gregor (gregod_at_[hidden])
Date: 2003-10-30 10:21:00
On Thursday 30 October 2003 02:18 am, Vladimir Prus wrote:
> Douglas Gregor wrote:
> > On Tuesday 28 October 2003 07:41 am, Vladimir Prus wrote:
> >> I'm about to make "stage" traverse dependencies, as was requested some
> >> time ago. The proposed semantic is described below. If anybody sees any
> >> problem with it, I'd be gratefull to know.
> > Thank you! BoostBook will be forever improved.
> Do you mean you'll be able to move generated docs into "html" directory
> without black magic. Oh, I did not even though this change will help
That's possible, too, but I was thinking about the Doxygen transformation.
Right now, it's done like this:
* Doxygen takes a bunch of headers and produces a bunch of XML files
* XSLT processor takes those XML files and combines them into one big file
* XSLT processor takes that one big file and transforms it into BoostBook
* The BoostBook XML is then staged over to the documentation directory
* BoostBook processing (XSLT) then starts at the top-level BoostBook
document, and builds DocBook/HTML/etc.
The staging is really really important, because we want to check the staged
version into CVS. If the user has Doxygen configured, the BoostBook module
sets up all of the dependencies in that change and everything magically
works; if the user does NOT have Doxygen configured, BoostBook skips the
Doxygen part and starts the dependency at the staged copy.
At least, that's how it's going to work, but staging didn't allow dependencies
(until now!). All that's left is the is-it-or-isn't-it-there logic (should be
easy). This really solves a major conceptual problem with BoostBook's
transformation: how to support multiple tools generating BoostBook input
without requiring everyone to have all of those tools available.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk