Boost logo

Boost-Build :

From: K. Noel Belcourt (kbelco_at_[hidden])
Date: 2006-10-11 18:15:23

On Oct 11, 2006, at 1:45 AM, Vladimir Prus wrote:

> On Wednesday 11 October 2006 01:23, K. Noel Belcourt wrote:
>> Is there a way to output the full, unique set of source
> Yes. Take a look at the 'gettext.jam:update-translations-
>'. It
> collects all source files.
>> and header
>> files necessary to build a target?
> No, since header scanning is done only when building targets.
>> If there was, I could take the
>> list of files and pipe it to tar.
>> Any thoughts on how to go about this?
> Just to check -- is there a reason while wholesale 'tar' of a
> directory is not
> OK?

In our case, the source trees are littered with unused files, build
system files, multiple copies of the source, tests, documentation,
not to mention people who have built in the source tree. Our current
tools are rather poorly designed and permitted (actually required)
building in the source tree. I've produced a new build system based
on bjam, but while we test it and implement some additional
capabilities (like packaging the source for distribution), I can't
break the existing system by removing any of these files.

Besides, I think it's a much more elegant solution to use the
dependency tree to determine what is needed to build a given product
from scratch. Using bjam to generate the full source distribution
and then using bjam to build from that distribution is where I want
to go.

> If there is, I'm afraid you'd need to tar "complete" set of headers
> anyway, and use the logic similar to gettext generator to get the
> list of
> sources.

The gettext generator looks about right. The only thing I'd like to
avoid is having to actually build a target in order to generate the
complete set of headers and sources for a target. I think I know how
to accomplish this but I'll give a few things a try before asking
about that.


-- Noel

Boost-Build list run by bdawes at, david.abrahams at, gregod at, cpdaniel at, john at