[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20151022212650.GF10551@port70.net>
Date: Thu, 22 Oct 2015 23:26:51 +0200
From: Szabolcs Nagy <nsz@...t70.net>
To: musl@...ts.openwall.com
Cc: Rich Felker <dalias@...c.org>
Subject: Re: __write alias missing on arm platforms
* Sebastian Gottschall <s.gottschall@...wrt.com> [2015-10-22 23:18:54 +0200]:
> Am 22.10.2015 um 22:55 schrieb Rich Felker:
> >On Thu, Oct 22, 2015 at 10:21:54PM +0200, Sebastian Gottschall wrote:
> >>musl is missing the __write function which seem to be a alias to "write".
> >>this is required by libgcc on arm platforms. under some conditions
> >>the compiler will generate code which does reference
> >>this function then which leads to unresolved symbols
> >>(seen on squid 4.0.1 on xscale)
> >>
> >>/home/xfs/toolchains/toolchain-armeb_xscale_gcc-5.2.0_musl-1.1.11/bin/../lib64/gcc/armeb-openwrt-linux-muslgnueabi/5.2.0/libgcc.a(linux-atomic-64bit.o):
> >>In function `__check_for_sync8_kernelhelper':
> >>/home/xfs/openwrt/trunk2/trunk/build_dir/toolchain-armeb_xscale_gcc-5.2.0_musl-1.1.11/gcc-5.2.0/libgcc/config/arm/linux-atomic-64bit.c:59:
> >>undefined reference to `__write'
> >>collect2: error: ld returned 1 exit status
> >>l
> >__write is not a public API even in glibc as far as I can tell. This
> >looks like an abuse of an accidentally-public symbol by libgcc. For
> >musl target anyway it should just be patched out of libgcc, e.g. with
> >#ifdef.
> you mean i have to send a bug report to GCC maintainers? i its part if the
> libgcc sourcecode and is not provided by libgcc, but just referenced
> i checked uclibc and glibc. both define __write as alias for write just
> because of libgcc
>
> so lets declare musl as broken on arm if gcc is used, if this is a solution
> for you :-)
>
i will submit a bug report about this.
it should be fixed in gcc, __write is not
part of any official abi (not even lsb!),
there is no reason to use that here,
it is broken on many platforms not just musl
(bionic does not have it either).
meanwhile we can think about some workaround
(the most obvious is to patch libgcc since
you are rebuilding gcc anyway, just remove
the __write call and leave abort there)
> >
> >Rich
> >
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
