Subject: Re: [Boost-build] [headers] some weird issues with hardlinks
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2015-10-10 12:23:58
On 10/10/2015 06:43 AM, Raffi Enficiaud wrote:
> Hi all,
> I have just deployed a private build agent for testing boost.test. I
> have a problem with one of the headers not showing up.
> The build I am making is just this
> - checkout boost[develop]
> - checkout boost.test[specificbranch]
> - run b2 headers
> - run b2 on boost.test regression
> - rmdir /s boost folder
> - rmdir /s bin.v2
> The trace of the error is shown below: the header
> "boost/utility/detail/result_of_iterate.hpp" is not there, only on Win7.
> It works well for the other targets (OSX + Linux).
> The funny thing is that when I run "b2 headers" from a command line, the
> header is in the right place.
> The questions are:
> - Am I doing things right? I thought that running b2 headers is
> sufficient for having everything under $BOOST_ROOT/boost in the right
> place. Am I missing something?
> - what makes b2 think that something has to be copied or not?
Whether files are copied and when they're copied
is determined by #include scanning. The reason
that this fails is probably because the file is being
#included through a macro, which Jam's regex based
scanner can't process.
> I can see
> in the logs hardlinks created "on demand", while I thought that running
> "b2 headers" should create all those links already.
Look at the log again. You're not
running b2 headers.
> - is there a way to avoid the creation of hard links (just a copy would
> do)? or to avoid the economic approach where they are created on demand?
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