From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Wed, 29 Jan 2014 09:30:13 +0100 Subject: [Buildroot] full rebuild and errors In-Reply-To: <52E8AA41.1020709@mind.be> References: <52E4FA8C.7060109@gmail.com> <52E6945C.40202@mind.be> <52E8120F.7080805@gmail.com> <20140128230904.1f01a5dc@skate> <52E8AA41.1020709@mind.be> Message-ID: <20140129093013.5ca63190@skate> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Dear Arnout Vandecappelle, On Wed, 29 Jan 2014 08:14:09 +0100, Arnout Vandecappelle wrote: > > No, it doesn't. The cache of ccache is located in > > ~/.buildroot-ccache/, because the intent of ccache is precisely to be > > able to re-use the cache contents across complete rebuilds, and even > > across multiple clones of Buildroot. > > But this is yet another proof that it doesn't work very well. And this > one really worries me, because it's a host build that failed. And for what it worth just two days ago, one of my colleague also had a weird compilation problem while building host-gettext (the link step was complaining because one of the object file it was given was built for i386, while the host machine is x86-64, and therefore the link was failing). The problem was reproducible even after a complete 'make clean all' cycle. And it turned out that clearing the cache solved the problem, so it was indeed a mis re-use of an existing object file. And it was also on the build of a host package, which is really weird. We know that our ccache handling is far from perfect in terms of detecting a change of the target compiler, but I wasn't aware of any theoretical problem regarding the build of host packages. Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com