From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Korsgaard Date: Sun, 15 Dec 2019 08:56:16 +0100 Subject: [Buildroot] [PATCH 1/1] package/python-subprocess32: add PYTHON_SUBPROCESS32_CONFIGURE_CMDS In-Reply-To: <20191214215428.75264bb3@windsurf.home> (Thomas Petazzoni's message of "Sat, 14 Dec 2019 21:54:28 +0100") References: <20191214171023.1650001-1-fontaine.fabrice@gmail.com> <20191214215428.75264bb3@windsurf.home> Message-ID: <875ziif2e7.fsf@dell.be.48ers.dk> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net >>>>> "Thomas" == Thomas Petazzoni writes: > On Sat, 14 Dec 2019 18:10:23 +0100 > Fabrice Fontaine wrote: >> Since commit 1745fcde740057951dcc5429f3bfabd103b764a1, >> python-subprocess32 fails to build because it runs configure with >> incorrect arguments so add a PYTHON_SUBPROCESS32_CONFIGURE_CMDS >> >> Fixes: >> - http://autobuild.buildroot.org/results/dcf944129392ee6cacc106e096d8d3adfa4447bb >> >> Signed-off-by: Fabrice Fontaine > When seeing this, I was a bit "meh, how did it work before?". So I > built python-subprocess32 with 1745fcde740057951dcc5429f3bfabd103b764a1 > reverted. What happens is that the ./configure script is executed, with > no arguments at all, and no CC variable is passed. So it happily uses > the host compiler, does a bunch of tests with it, and then uses the > resulting config.h file to build some target library. This is obviously > completely bogus! > With 1745fcde740057951dcc5429f3bfabd103b764a1, we now pass > CC=cross-compiler, so ./configure doesn't work because it uses a > cross-compiler, but no --host= value is passed, leading to the build > failure. > So 1745fcde740057951dcc5429f3bfabd103b764a1 really uncovered a real bug > in the python-subprocess32 packaging, which your patch fixes. Wow, (part of) this info would have been great to add to the commit message. -- Bye, Peter Korsgaard