Sorry - the %%PYTHON_SITELIBDIR%% I added at the end of the Makefile should be ${PYTHON_SITELIBDIR}
Update to graphics/opencolorio which fixes the gcc linking issues and documentation creation. Three questions - 1. The base code installs .so.1 and .so.1.0.7 and I added .so.1.0 - is the .so.1.0 needed/not-wanted/not-important? 2. To get the latex steps working I needed to rename .jpeg files to .jpg - this seems a wasted step - is there something I am missing or is this a quirk we work with? 3. With openimageio I needed to separate the python bindings into a separate port, which was initially due to dependency issues. While I don't see the dependency issue here, I do need to build the bindings for the doc creation step. Should I separate the python bindings install into a separate port or add an option to install them or is that my choice? Fix: Patch attached with submission follows:
This patch replaces the previous one. One small change added is to remove adding the rpath to LDFLAGS Also corrected the PYTHON_SITELIBDIR I pasted wrong before submission.
Responsible Changed From-To: freebsd-ports-bugs->mva I'll take it.
> Three questions - > 1. The base code installs .so.1 and .so.1.0.7 and I added .so.1.0 - is > the .so.1.0 needed/not-wanted/not-important? This is usually up to you, if there is no port or application explicitly requiring .so.1.0 - most should not, but stick with .so or .so.X. I would recommend not to create it to avoid unnecessary maintenance. > 2. To get the latex steps working I needed to rename .jpeg files to .jpg > - this seems a wasted step - is there something I am missing or is this > a quirk we work with? After looking at docs/CMakeLists.txt, it seems that the HTML version of the docs always gets build. Is a latex-based output necessary then? To lower the patching effort, I'd recommend to match CMakeLists.txt not to build the latex part by e.g. commenting out find_package(LATEX) and setting if(PDFLATEX_COMPILER) to e.g. if(FALSE) or if(0). > 3. With openimageio I needed to separate the python bindings into a > separate port, which was initially due to dependency issues. While I > don't see the dependency issue here, I do need to build the bindings for > the doc creation step. Should I separate the python bindings install > into a separate port or add an option to install them or is that my > choice? That's your choice. Cheers Marcus
This patch replaces the previous patch. Fixed gcc linking. Fixed documentation building. Removed utility apps - moving into own port (opencolorio-tools)
State Changed From-To: open->closed Committed, with minor changes. Thanks!