From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Korsgaard Date: Wed, 05 Jun 2013 20:52:49 +0200 Subject: [Buildroot] [autobuild.buildroot.net] Build results for 2013-06-04 In-Reply-To: <20130605184836.6ea2b205@skate> (Thomas Petazzoni's message of "Wed, 5 Jun 2013 18:48:36 +0200") References: <20130605063004.4425352C13E@lolut.humanoidz.org> <20130605182924.71a07679@skate> <20130605184836.6ea2b205@skate> Message-ID: <87txlccrum.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: Thomas> Hello, Thomas> On Wed, 5 Jun 2013 18:29:24 +0200, Thomas Petazzoni wrote: >> > host-libftdi-0.19 | 20 >> >> Those issues are caused by the bump of libtool, from the testing I >> could do. Reverting libtool to the previous version "solves" the >> problem. >> >> The thing that happens is that libftdi (for the target) has >> AUTORECONF=YES, but not host-libftdi. So libftdi gets autoreconfigured, >> and then when host-libftdi is being built, even though we don't have >> AUTORECONF=YES, when starting the build, it does execute >> 'CDPATH="${ZSH_VERSION+.}:" && cd . >> && /bin/bash /home/test/outputs/libftdi/build/host-libftdi-0.19/missing >> --run aclocal-1.11 ' and it ends up failing. Thomas> Ok, the reason it does execute "missing" and does a kind of automatic Thomas> autoreconf is because we have one patch that modifies configure.in. So Thomas> configure.in is more recent than the configure script itself. When Thomas> libtool/autoconf/automake has not been built before host-libftdi, then Thomas> this automatic autoreconf cannot proceed: That makes more sense. I didn't get how a target build (in a different directory) could affect it. So the fix is just HOST_LIBFTDI_AUTORECONF = YES? Thomas> and it in facts makes the thing work. Thomas> However, when this "automatic autoreconf" is done, and an Thomas> automake/autoconf/libtool set of tools, with the latest libtool, has Thomas> already been built and installed, then it generates some wrong stuff. Thomas> An easy fix is to add: Thomas> HOST_LIBFTDI_AUTORECONF = YES Thomas> to libftdi.mk. It doesn't really explain why it was working with Thomas> libtool 2.2 and no longer with libtool 2.4, but it works. Indeed, odd. Well, having _AUTORECONF = YES without HOST__AUTORECONF = YES is obviously wrong, so let's just fix that. -- Bye, Peter Korsgaard