2015-03-25 22:13 GMT+03:00 Ingo Feinerer <feine...@logic.at>:
> Dear OpenBSD useRs,
>
> please consider the following update of math/R to 3.1.3.
>
> Although this update is minor (3.1.2 -> 3.1.3) the diff is quite large
> as I used the opportunity to get rid of as many patches as possible. The
> plan is to make future updates (like to 3.2.0 which is scheduled for
> 2015-04-16) a lot easier.
>
> The current proposed diff builds fine for me on amd64 and passes the
> same regression tests as 3.1.2 with all patches did. Moreover, all my
> private R scripts run without problems with this new patch-reduced
> version.
>
> Please find below a detailed list of comments for each modified/deleted
> patch. I appreciate all comments.
>
> - Deleted patch-Makeconf_in
> Replaced by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> - patch-Makefile_in
> Build and install PDFs; works with BUILD_DEPENDS=print/texlive/base
>
> - patch-configure
> -Wl,-export-dynamic appears to work as well
> DYLIB_EXT correctly set by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> - Deleted patch-doc_Makefile_in
> Build NEWS.pdf; works with BUILD_DEPENDS=print/texlive/base
>
> - Deleted patch-etc_Makeconf_in
> Replaced by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> - Deleted patch-src_appl_Makefile_in
> Replaced by MAKE_FLAGS += Rblas_la="libRblas.so.${LIBRblas_VERSION}"
>
> - Deleted patch-src_extra_blas_Makefile_in
> Replaced by MAKE_FLAGS += Rblas_la="libRblas.so.${LIBRblas_VERSION}"
>
> - Deleted patch-src_include_R_ext_eventloop_h
> Not needed; works without
>
> - Deleted patch-src_main_CommandLineArgs_c
> 10 years old patch, unchanged for 10 years upstream, just syntactic sugar
>
> - patch-src_main_Makefile_in
> DYLIB_EXT correctly set by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> - Deleted patch-src_main_connections_c
> Not strictly needed. This patches replaces one vsprintf with vsnprintf
> but there are many similar cases that could be addressed. So skip this
> one as well to minimize differences.
> For the signed char c vs. char c I see no difference in this use case.
> But maybe because of my amd64 platform.
>
> - Deleted patch-src_main_gram_y
> Again signed char c vs. char c
>
> - Deleted patch-src_main_main_c
> I was not able to reproduce the warning "failed to set alternate
> signal stack"
>
> - Deleted patch-src_main_platform_c
> The port also builds without this patch with USE_SYSTRACE=Yes for me
>
> - Deleted patch-src_modules_lapack_Makefile_in
> Replaced by MAKE_FLAGS += Rlapack_la="libRlapack.so.$(LIBRlapack_VERSION)"
>
> - Deleted patch-src_nmath_standalone_Makefile_in
> DYLIB_EXT correctly set by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> - Deleted patch-src_scripts_config
> DYLIB_EXT correctly set by MAKE_FLAGS += DYLIB_EXT=".so.${LIBR_VERSION}"
>
> Moreover, I sent patch-src_main_Rstrptime_h upstream. Let's see if it
> gets processed and accepted.
Please, don't remove patches that allow to avoid texlive dependency.
They do help heavily in bulk builds. The best would be to find way to
be able do it either way - by picking up automatically during
configure state, or forcing state with configure switches/environment.
If that's too hard or won't be accepted upstream, then those patches
should stay in.
--
WBR,
Vadim Zhukov
