Re: [Boost-testing] process_jam_log build: missing boost/config/compiler/visualc.hpp

Subject: Re: [Boost-testing] process_jam_log build: missing boost/config/compiler/visualc.hpp
From: Jim Bell (Jim_at_[hidden])
Date: 2014-01-19 23:12:36

On 2014-01-17 2:02 PM, Jim Bell wrote:
>
> On 2014-01-17 10:09 AM, Beman Dawes wrote:
>> On Wed, Jan 15, 2014 at 7:39 AM, Jim Bell <Jim_at_[hidden] 
>> <mailto:Jim_at_[hidden]>> wrote:
>>
>>     It looks like parts of boost/config are missing:
>>
>>     ..\..\..\boost/config.hpp(35) : fatal error C1083: Cannot open
>>     include file: 'boost/config/compiler/visualc.hpp': No such file
>>     or directory
>>
>>     It was also missing boost/config/user.hpp, but I added an empty one.
>>
>>     boost/config/compiler (as pulled from git in the regression test)
>>     has just one file: nvcc.hpp.
>>
>>
>> Rene pushed up a change 11 hours ago that should have fixed missing 
>> header problems for all platforms. If you run into further problems, 
>> please report them ASAP.
>
> No change in behavior from a run script downloaded today, 1pm (1 hour 
> ago). I removed my empty user.hpp, and that's where it's stuck now.
>
> ...
>

Comparing my working Linux regression to my Win32, it appears that the 
hardlinks aren't being set up correctly. I see a symbolic link from 
boost_root/boost/config -> ../libs/config/include/boost/config, but no 
such linkage on the Win32 side. libs/config/include/boost/config 
directory is populated correctly (on Win32 also).

Here's more output, having run with a clean slate. Note all the hardlink 
creations, but I don't see boost/config in there anywhere. 
http://pastebin.com/rQAVuhEM

Is it doing a simple copy on win32 instead of a true hardlink? Something 
like SysInternals' junction?