From: Peter Korsgaard Date: Thu, 20 Feb 2014 08:49:31 +0100 Subject: [Buildroot] [PATCH v2 1/1] icu: detect and add compiler symbol prefix to the assembly code In-Reply-To: <20140220083102.609c5783@skate> (Thomas Petazzoni's message of "Thu, 20 Feb 2014 08:31:02 +0100") References: <1391519206-17889-1-git-send-email-rjbarnet@rockwellcollins.com> <87wqgqyaxy.fsf@dell.be.48ers.dk> <20140220083102.609c5783@skate> Message-ID: <87sirexlys.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: > Dear Peter Korsgaard, > On Wed, 19 Feb 2014 23:50:01 +0100, Peter Korsgaard wrote: >> We recently added this patch to icu for bfin symbol prefix, but from >> todays autobuilder I see that icu's configure script completely doesn't >> support uclinux (E.G. bfin): >> >> http://autobuild.buildroot.net/results/d11/d118b4f753ade63201758e1b14ca54d96ec77698/build-end.log >> >> What's going on here? Should I revert this patch and instead mark icu as >> unavailable on bfin? > No. I believe icu support bfin-linux-uclibc (i.e FDPIC), but not > bfin-uclinux. I've modified icu so that it accepts bfin-uclinux as > well, but apparently icu generates an ELF file by itself (!), and puts > it in an .a archive. Which of course is not really to the taste of the > FLAT format ld used on bfin-uclinux. Ahh yes, I missed the different bfin variants. > I'm trying to figure out if icu can be convinced to generate source > code instead and compile it, but the icu source code is weird, to say > the least. > So I believe I will send a patch that disables ICU when !BR2_BINFMT_ELF. Ok, thanks. -- Bye, Peter Korsgaard