From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Tue, 18 Sep 2018 09:51:08 +0200 Subject: [Buildroot] [PATCH v2 3/3] package/piglit: new package In-Reply-To: <84ded507-27f9-9e33-a116-becc8dda57f0@smile.fr> References: <20180211162136.15605-1-romain.naour@gmail.com> <20180211162136.15605-3-romain.naour@gmail.com> <20180916233840.79495eba@windsurf> <84ded507-27f9-9e33-a116-becc8dda57f0@smile.fr> Message-ID: <20180918095108.1a128bcf@windsurf> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hello, On Mon, 17 Sep 2018 22:10:02 +0200, Romain Naour wrote: > > This is surprising, what does it do on the host during the build > > process that requires those tools ? Or is it that piglit generates a > > report on the target, and then provides a bunch of host tool to process > > such reports ? > > Reports can be processed on the target directly. > I believe it's for the same reason that we need host-numpy. What "same reason" ? > I think the build system is checking the host-python instead of python for the > target. > > I'm not sure how to fix this... So you mean that all those Python packages are not needed on the host, but because the build system is bogus and checks their presence on the host, we have to have them as dependencies ? So indeed, there's a macro in cmake/Modules/PythonModule.cmake that is used to check for Mako, Numpy and Six, and it works by running the host Python. I imagine they are in fact not needed on the host, and this macro needs to be adjusted. However, I'm not sure what's the right way to test if a *target* Python module is available. Maybe Yegor can help on this ? :-) Thomas -- Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com