ELIXIR_LIB_ROOT is the default Elixir library path in LOCALBASE.
If MIX_REWRITE is defined, all optional dependencies will be stripped
out (test, doc, etc.), while all required libs will be converted to
actual code paths in ELIXIR_LIB_ROOT.
For example, it will turn the following dependency definitions:
defp deps do
[
{ :inflex, "~> 1.0" },
{ :estree, "~> 2.0" },
{ :shouldi, only: :test },
{ :earmark, "~> 0.1", only: :dev },
{ :ex_doc, "~> 0.7", only: :dev }
]
end
into these:
defp deps do
[
{ :inflex, path: "/usr/local/lib/elixir/lib/inflex", compile: false },
{ :estree, path: "/usr/local/lib/elixir/lib/estree", compile: false },
]
end
Setting MIX_REWRITE allows escriptize to bundle all dependencies in the
script, else it will not be able to pull them from the usual code path.
It already works for all Elixir ports and should make patches to mix.exs
unnecessary, but for now the default is off.
Sneak in a whitespace fix while I'm here.
The transition from gcc-aux to gcc5-aux in the Ada framework has been
blocked by the inability to build gtkada3 and, once resolved, GPS (due
to tight locking with compiler).
A few days ago, Adacore made their annual release of their main libre
products, include GPS. However, some products were tightly coupled with
the recent compilers, so in order to upgrade, the compiler had to be
switched and dependencies require many ports to be upgraded at once:
* lang/asis
* devel/gnatcoll
* devel/gps
* x11-toolkits/gtkada3
* www/aws
* www/aws-demos
While the version upgrades were modest in most cases (gps, gtkada3), the
amount of work put into each port was significant. There are too many
improvements to mention here. A few include the removal of dynamic
package lists and incorporating gnatcoll into gps to avoid building it
twice. A private "exp-run" was done all on all 50+ Ada ports to ensure
they still build.
Also, a new argument was added to Uses/ada.mk, "run", that pulls in the
GNAT compiler as a run depends. This was necessary for GPS that will
not launch correctly without the compiler in place.
for CMake based ports. By default CMake picks up the highest
available version of Python package, therefore ports which use CMake
may fail to build or link to different Python library than one could
expect (e.g. bugs 199685, 200518).
In essence this commit reverts r366996.
PR: 168159
Exp-run by: antoine
Approved by: mva
Make output messaging clearer about what uniquefiles is doing
especially with regard to moving and linking (default links).
- Add "action" prefix (Move, Link) to operations performed
- Use the form "Creating unique files" for the stage summary
- Be clear that we're processing files that *need* prefixes or
suffixes, and not moving files that *have* prefixes or suffixes"
Differential Revision: https://reviews.freebsd.org/D2706
Reviewed by: mat
Approved by: mat (portmgr)
conversion in r364450. This feature is still very useful for meta packages
to be able to easily depend on py-* packages using the correct PY_* vars.
For the sake of POLA retain the same name so previously working ports just work.
Add an XFCE SITE_SUBDIR abbreviation and use it forf or the xfce ports,
which removes the need to set MASTER_SITE_SUBDIR in any of them.
This fixes ports that have USES=xfce but do not use the XFCE MASTER_SITE,
namely sysutils/xfce4-bsdcpufreq-plugin.
With hat: portmgr
This fixes building devel/gettext-runtime with gcc from lang/gcc*.
With base system compiler gettext includes /usr/include/iconv.h but
lang/gcc* implies -I/usr/local/include and gettext calls
libiconv_set_relocation_prefix when it sees _LIBICONV_VERSION which
isn't implemented (and isn't needed) by libc iconv.
Reported by: Dewayne Geraghty <dewayne.geraghty@heuristicsystems.com.au>
- Remove const qualifier from iconv(3) to match POSIX:
http://pubs.opengroup.org/onlinepubs/9699919799/functions/iconv.html
- Patch iconv.h to expose more GNU extensions when LIBICONV_PLUG is
defined because the base system iconv supports these extensions too.
Add/remove patches to/from ports to call iconv with non-const arguments.
This breaks some ports on FreeBSD 10 because base system iconv.h still has
the const qualifier. Fix this by letting USES=iconv add a build dependency
on converters/libiconv so ports can use its iconv.h (with LIBICONV_PLUG
defined) instead of the base system iconv.h.
This exposed some ports that link with libiconv when it is available instead
of using libc iconv. In these cases one of the following changes has been
made:
- patch configure scripts to test for libc iconv first
- add ac_cv_lib_iconv_libiconv=no or similar to CONFIGURE_ARGS to disable
some configure tests
- converters/wkhtmltopdf: this includes Qt4 so add a patch from devel/qt4
- lang/gcc5-aux: respect CFLAGS and friends during configure such that
LIBICONV_PLUG is defined in the iconv test, also switch to external
gettext
- mail/gnarwl: replace patches with CPPFLAGS/LIBS
- multimedia/ffmpeg2theora: remove iconv test from SConstruct and use
ICONV_LIB in port Makefile instead, also fix a bug in subtitles.c
- net-im/licq: finish conversion to cmake
- net-mgmt/bandwidthd, net-mgmt/icinga, net-mgmt/nagios, net-mgmt/nagios4:
don't need iconv
- textproc/p5-XML-TinyXML: finish conversion to USES=perl5
Other changes:
- databases/qdbm and slaves: respect CFLAGS and friends, also enable bzip2
and lzo support
- games/ldmud: respect CFLAGS and friends
- graphics/inventor: replace some patches with MAKE_ARGS/MAKE_ENV to respect
CFLAGS and friends, also remove FreeBSD/alpha patch and add missing xorg
dependencies
PR: 199099
Exp-run by: antoine
Approved by: portmgr (antoine)
- Override it for USES=ninja
With this commit and r383571, ports using ninja and waf now respect
MAKE_JOBS_NUMBER when it's equal to 1
PR: 197910
With hat: portmgr
Simplify gnustep ports
Hook into the regular ports framework:
- LIB_DEPENDS for library dependencies
- Use regular USE_LDCONFIG
Reuse USES=objc (automatic)
USE_GNUSTEP is now a macro to set the dependencies and build feature needed.
Accepted arguments: back base build gui
Merge deskutils/preferencepanes into deskutils/systempreferences
Plug waf into MAKE_CMD and CONFIGURE_CMD so the regular defined targets can be
reused
Always define _MAKE_JOBS so that when bsd.port.mk will stop overwritting
_MAKE_JOBS when parallel jobs are disabled we can enforce -j1 (which is needed
to really disable parallelisation with waf
WAF_CMD has been created to allow one to override the location of the waf script
relatively to WRKSRC
CONFIGURE_TARGET is by default defined to "configure"
ALL_TARGET is by default defined to "build"
INSTALL_TARGET is by default defined to "install"
USES=waf is by default stagedir safe
