From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Tue, 23 Feb 2016 23:35:28 +0100 Subject: [Buildroot] [PATCH v4] package/python-pillow: new package In-Reply-To: References: <1455962149-29459-1-git-send-email-angelo.compagnucci@gmail.com> <20160221152621.52c42acd@free-electrons.com> Message-ID: <20160223233528.4016ea2b@free-electrons.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hello, On Tue, 23 Feb 2016 23:18:47 +0100, Angelo Compagnucci wrote: > Right. > > I actually implemented something lie this in pkg-python.mk > > [...] > $(2)_BUILD_TARGET ?= build > [...] > $(2)_BASE_BUILD_TGT = $$($(2)_BUILD_TARGET) > [...] > > and seems to work and not breaking other packages. > This way I can add: > > PYTHON_PILLOW_BUILD_TARGET = build_ext > > What do you think? Since it's the first package to require this, I would rather suggest to override PYTHON_PILLOW_BUILD_CMDS in python-pillow.mk, so that this hack is limited to this package. Should more packages need this in the future, we can add better support in the infrastructure. But a Python package that needs to pass options only to build_ext is IMO broken, and should allow them to be passed to the "build" target. build_ext is more or less an internal step, which we shouldn't have to worry about. > >> +Signed-off-by: Angelo Compagnucci > > > > Can you submit this patch upstream or at least report the issue? > > I don't think so. That piece of code is used to guess the platform on > which your are compiling with the assumption that host arch == target > arch. Probably, pillow is not originally designed to be cross compiled > and the build system would require a rewrite for a more structured > approach. > The only way to let buildroot passing the appropriate options to build > and setup is to remove that piece of code. Well, you can make it upstreamable by making it understand some environment variable, or better some option, to indicate that we are cross-compiling. So yes, your patch is not acceptable upstream as is, but my comment is precisely that is should ideally be in a form that can potentially be accepted upstream. Thanks! Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com