other than the usual "python3/<blank>" python version selection and
remove setting MODPY_VERSION=${MODPY_DEFAULT_VERSION_3} again from the
affected ports.
if a port needs 2.x then set MODPY_VERSION=${MODPY_DEFAULT_VERSION_2}.
This commit doesn't change any versions currently used; it may be that
some ports have MODPY_DEFAULT_VERSION_2 but don't require it, those
should be cleaned up in the course of updating ports where possible.
Python module ports providing py3-* packages should still use
FLAVOR=python3 so that we don't have a mixture of dependencies some
using ${MODPY_FLAVOR} and others not.
While waiting for this to appear in a proper Python 3.9.x release,
use upstream's commit for this severe sprintf bug.
The bug was reported on Jan 16 and the fix was available since Jan 18,
but only 3.6 and 3.7 have new releases as of now.
ok sthen
While waiting for this to appear in a proper Python 3.8.x release,
use upstream's commit for this severe sprintf bug.
The bug was reported on Jan 16 and the fix was available since Jan 18,
but only 3.6 and 3.7 have new releases as of now.
ok sthen
MODPY_BIN_ADJ's perl snippt accepts multiple files so save a few execs.
To make it more robust, also append `--' to it such that ports cannot
(accidentially) pass options; I've checked the tree that no port does
this on purpose.
The only case where this could fail is with huge MODPY_ADJ_FILES but
that is not the case in our tree; ports where lots of shebangs are
fixed have their own construct around it, e.g. textproc/calibre which
uses the `find -exec ${MODPY_ADJ_FILES} {} +' idiom.
OK sthen
setuptools picks up plugins during the configure stage whether a port
needs them or not. In some cases (currently just setuptools_scm), these
register hooks to run in the install stage; if the plugin is not a
listed build dependency this will cause the install stage to fail.
Recently reported by naddy with py-sphinx but we've seen spurious
failures elsewhere before which are likely due to this.
ok kmos@
Some OCaml ports build native-code programs on architectures where
supported, most of them probably don't need the big ocaml package at
runtime (for example net/unison). This commit introduces
MODOCAML_RUNDEP, setting it to `if-not-native' should do the right
thing.
Earlier diffs by bket@. ok sthen@ bket@
egcc -static was broken on most arches; the static binary wrongly
requested ld.so(1), but /usr/lib/rcrt0.o wants to do the relocation
without ld.so(1). The fix is to drop ld.so,
%{!static:-dynamic-linker /usr/libexec/ld.so}
Remove "%{!dynamic-linker:...}" because gcc -dynamic-linker is not an
option. Correct "%{!r*:...}" to "%{!r:...}" because gcc -r doesn't
take an option argument. Use "%{!nostdlib:-L/usr/lib}" on lld
arches, like base-gcc.
Edit the powerpc LINK_SPEC in rs6000/openbsd.h to be like the others.
Add "%{!nostdlib:-L/usr/lib}" so I can try lld on powerpc. For now,
don't edit the powerpc64 LINK_SPEC.
ok pascal@ (maintainer)
