[Bf-cycles] Building OSL for Blender

Sun Feb 1 23:12:43 CET 2015

Hi Sergey, many thanks for that.  I'll try building the upstream branch as
you suggested, and the bisection process.  One quick thing -- I was trying
to build by running cmake, and pointing it to all the pre-built mac
libraries that I downloaded from blender.org.  Presumably this is what you
do?  The problems I'm having seem to be related to static linking, and I
notice most of the prebuilt libs are static libraries.

On Sun, Feb 1, 2015 at 4:59 PM, Sergey Sharybin <sergey.vfx at gmail.com>
wrote:

> Linux build environment uses gcc-4.7 (which could make affect on the
> linking issue you're having). I also just tested with clang-3.5 here and it
> worked just fine. So afraid not much help can provide you, however:
>
> - Try building upstream version 1.5.11 and see if it works fine (1.5.11 is
> what the branch is based on, plus some commits on top of it). If it works
> then you'll narrow possible build regressions form our side to 3 commits.
> It if fails. If it doesn't compile:
> - Try building latest OSL upstream version
>
> If latest upstream compiles fine then you can start bisecting for the
> commit which solves compilation for you. If it fails then we're unable to
> help you and you need to contact OSL guys in their ML.
>
> On Sat, Jan 31, 2015 at 5:05 PM, Kevin Atkinson <kevin.atkinson at gmail.com>
> wrote:
>
>> Hi Sergey,
>>
>> Thanks for the reply.
>>
>> >Is it OSL which has issues with compilation or it's blender which fails
>> to link against OSL library you built?
>>
>> It's compilation of the patched version of OSL that is used in Blender.
>> The libraries actually compile without errors, but anything that links
>> against them (e.g., the test executables in OSL or Blender) will have
>> missing symbols.  All the missing symbols are from the C++ standard
>> library.  So it's lots of this:
>>
>> *Linking CXX executable accum_test*
>>
>> *Undefined symbols for architecture x86_64:*
>>
>> *  "std::string::find_first_of(char const*, unsigned long, unsigned long)
>> const", referenced from:*
>>
>> *      OpenImageIO::v1_4::Strutil::get_rest_arguments(std::string const&,
>> std::string&, std::map<std::string, std::string, std::less<std::string>,
>> std::allocator<std::pair<std::string const, std::string> > >&) in
>> libOpenImageIO.a(strutil.cpp.o)*
>>
>> *  "std::string::find(char, unsigned long) const", referenced from:*
>>
>> *      bool boost::char_separator<char, std::char_traits<char>
>> >::operator()<__gnu_cxx::__normal_iterator<char const*, std::string>,
>> std::string>(__gnu_cxx::__normal_iterator<char const*, std::string>&,
>> __gnu_cxx::__normal_iterator<char const*, std::string>, std::string&) in
>> libOpenImageIO.a(strutil.cpp.o)*
>>
>> *  "std::basic_stringbuf<char, std::char_traits<char>,
>> std::allocator<char> >::str() const", referenced from:*
>>
>> *      OpenImageIO::v1_4::ustring::getstats(bool) in
>> libOpenImageIO.a(ustring.cpp.o)*
>>
>> *
>> OpenImageIO::v1_4::Strutil::wordwrap(OpenImageIO::v1_4::string_view, int,
>> int) in libOpenImageIO.a(strutil.cpp.o)*
>>
>> *      std::string OpenImageIO::v1_4::Strutil::format<long long>(char
>> const*, long long const&) in libOpenImageIO.a(strutil.cpp.o)*
>>
>> *      std::string OpenImageIO::v1_4::Strutil::format<int, double, char
>> const*>(char const*, int const&, double const&, char const* const&) in
>> libOpenImageIO.a(strutil.cpp.o)*
>>
>> *      std::string OpenImageIO::v1_4::Strutil::format<int, int>(char
>> const*, int const&, int const&) in libOpenImageIO.a(strutil.cpp.o)*
>>
>> *      std::string OpenImageIO::v1_4::Strutil::format<int>(char const*,
>> int const&) in libOpenImageIO.a(strutil.cpp.o)*
>>
>> *      std::string OpenImageIO::v1_4::Strutil::format<int, int,
>> double>(char const*, int const&, int const&, double const&) in
>> libOpenImageIO.a(strutil.cpp.o)*
>>
>>       ...
>>
>> All the libraries that I link OSL to are the prebuilt ones for mac in
>> darwin-9.x.universal.  I'm using clang 6.0 on Mavericks.
>>
>>
>>
>> On Sat, Jan 31, 2015 at 5:23 AM, Sergey Sharybin <sergey.vfx at gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> That's a bit vague information. Is it OSL which has issues with
>>> compilation or it's blender which fails to link against OSL library you
>>> built?
>>>
>>> In any case, having an exact error messages would help.
>>>
>>> On Sat, Jan 31, 2015 at 7:41 AM, Kevin Atkinson <
>>> kevin.atkinson at gmail.com> wrote:
>>>
>>>> I'm using OSL gabor noise for displacements in Blender, and I've been
>>>> noticing nasty circle artifacts, coming from the rather coarse thresholding
>>>> of the gaussian kernel.  I've been doing experiments with the gabor noise
>>>> code outside of blender, and got very good results by simply changing the
>>>> line
>>>>
>>>> static const float Gabor_Truncate = 0.02f;
>>>>
>>>>
>>>> to
>>>>
>>>>
>>>> static const float Gabor_Truncate = 0.0002f;
>>>>
>>>>
>>>> This causes the search radius for impulses to be about half again as large, at some performance cost.  But not bad for reducing the artifact magnitude by a factor of a hundred.
>>>>
>>>>
>>>> I'd like to get this working in Blender just for my own use and further testing, but I'm having trouble building OSL for blender.  I've checked out the "blender-fixes" branch, but I get linking errors because for some reason it doesn't link to the C++ standard library.
>>>>
>>>>
>>>> I don't know much about CMake, and I am under the gun a bit, so I'm trying to to get in touch with whoever built the mac OSL library for blender to see if they have any suggestions.  Would that be Lukas Tönne?
>>>>
>>>>
>>>> Many thanks, and any help much appreciated,
>>>>
>>>>
>>>> Kevin Atkinson
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Bf-cycles mailing list
>>>> Bf-cycles at blender.org
>>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>>
>>>>
>>>
>>>
>>> --
>>> With best regards, Sergey Sharybin
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>
>
> --
> With best regards, Sergey Sharybin
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150201/f1970e3a/attachment.htm 