Hello.
On 07/16/2018 01:33 AM, Simon Glass wrote:
Hi,
On 12 July 2018 at 18:54, David Gibson <david@xxxxxxxxxxxxxxxxxxxxx> wrote:
On Thu, Jul 12, 2018 at 04:10:41PM +0200, frenzy@xxxxxxxxx wrote:
From: Lumir Balhar <lbalhar@xxxxxxxxxx>
Python 2 is still the default but it can be changed by
setting environment variable PYTHON before build/test.
Signed-off-by: Lumir Balhar <lbalhar@xxxxxxxxxx>
As I noted on an earlier version, I'd like to get Simon Glass's review
before merging these, since the Python bindings are his work.
However, a few preliminary comments from me.
---
Makefile | 3 ++-
pylibfdt/Makefile.pylibfdt | 4 ++--
tests/Makefile.tests | 6 +++---
tests/run_tests.sh | 2 +-
4 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/Makefile b/Makefile
index 6d55e13..c84eb66 100644
--- a/Makefile
+++ b/Makefile
@@ -24,6 +24,7 @@ BISON = bison
LEX = flex
SWIG = swig
PKG_CONFIG ?= pkg-config
+PYTHON ?= python2
INSTALL = /usr/bin/install
INSTALL_PROGRAM = $(INSTALL)
@@ -133,7 +134,7 @@ all: $(BIN) libfdt
# We need both Python and swig to build/install pylibfdt.
# This builds the given make ${target} if those deps are found.
check_python_deps = \
- if $(PKG_CONFIG) --cflags python2 >/dev/null 2>&1; then \
+ if $(PKG_CONFIG) --cflags $(PYTHON) >/dev/null 2>&1; then \
if which swig >/dev/null 2>&1; then \
can_build=yes; \
fi; \
diff --git a/pylibfdt/Makefile.pylibfdt b/pylibfdt/Makefile.pylibfdt
index 9507d3d..a05984c 100644
--- a/pylibfdt/Makefile.pylibfdt
+++ b/pylibfdt/Makefile.pylibfdt
@@ -8,13 +8,13 @@ PYMODULE = $(PYLIBFDT_objdir)/_libfdt.so
define run_setup
SOURCES="$(1)" CPPFLAGS="$(CPPFLAGS)" OBJDIR="$(PYLIBFDT_objdir)"
VERSION="$(dtc_version)"
- $(PYLIBFDT_objdir)/setup.py --quiet $(2)
+ $(PYTHON) $(PYLIBFDT_objdir)/setup.py --quiet $(2)
endef
Using the environment variable to select the default python for
install is fine, but..
$(PYMODULE): $(PYLIBFDT_srcs)
@$(VECHO) PYMOD $@
$(call run_setup, $^, build_ext --inplace)
- mv _libfdt.so $@
+ mv _libfdt.*so $@
install_pylibfdt: $(PYMODULE)
$(VECHO) INSTALL-PYLIB; \
diff --git a/tests/Makefile.tests b/tests/Makefile.tests
index 2c2c4fd..554c840 100644
--- a/tests/Makefile.tests
+++ b/tests/Makefile.tests
@@ -76,13 +76,13 @@ tests_clean:
rm -f $(TESTS_CLEANFILES)
check: tests ${TESTS_BIN} $(TESTS_PYLIBFDT)
- cd $(TESTS_PREFIX); ./run_tests.sh
+ cd $(TESTS_PREFIX); PYTHON=$(PYTHON) ./run_tests.sh
checkm: tests ${TESTS_BIN} $(TESTS_PYLIBFDT)
- cd $(TESTS_PREFIX); ./run_tests.sh -m 2>&1 | tee vglog.$$$$
+ cd $(TESTS_PREFIX); PYTHON=$(PYTHON) ./run_tests.sh -m 2>&1 | tee vglog.$$$$
..for testing, I'd be much happied if we ran the pylibfdt tests under
*both* Python2 and Python3 if available. Makes it much likelier that
we'll catch breakages for one or the other early.
I definitely agree with that, it is pretty important.
I am not sure how this can be done without breaking backward
compatibility. I have to say that I am not dtc nor libfdt expert so I
might be missing something.
AFAIK it's not possible to build the extension module for both Pythons
in the same time. In Makefile.pylibfdt on line 17, there is a `mv`
command which renames built extension to standard name
pylibfdt/_libfdt.so. When we build it again with a different Python
version, the previous one will be replaced here.
I can try to prepare some code which will handle whether use Python 2 or
Python 3 (or both) for testing based on extension file name but it will
mean that we cannot rename built extension file.
Other than that it looks fine.
Regards,
Simon
--
To unsubscribe from this list: send the line "unsubscribe devicetree-compiler" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
