Subject: Re: [boost] Dependency report
From: Edward Diener (eldiener_at_[hidden])
Date: 2017-07-30 22:08:59
On 7/30/2017 8:21 AM, Raffi Enficiaud via Boost wrote:
> Le 30.07.17 à 10:51, Peter Dimov via Boost a écrit :
>> Raffi Enficiaud wrote:
>>> For some libraries that are using boost.test, it seems that boost.test
>>> is included in the weight. But this is in fact not totally true: as it
>> is shown, it is a "developer" weight that wants to run the tests and
>> compile the doc.
>> The dependencies in the report are computed based on the headers in
>> `include`. If one of them includes a Boost.Test header, Test is
>> considered a dependency.
>> The usual practice is to move such private headers to the `test`
> That is weird though that eg. boost.graph depends on boost.test. After
> your reply, it is indeed in <boost/graph/graph_test.hpp> which should
> then be indeed moved somewhere else.
I agree with you. The <boost/graph/graph_test.hpp> is only used by the
graph tests and should be part of the graph test directory and not the
include directory path. It is not in any way part of the actual header
>>> Would it be somehow possible to account only for the source code that
>>> generates the libraries, as seen by users?
>> This would imply that header-only libraries have no dependencies, which
>> is absurd.
> You got me wrong here.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk