[Bug dynamic-link/15006] Libraries in ld.so.cache ignored by ld-linux-armhf.so.3 on armv6l

["steve.mcintyre at linaro dot org" <sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=15006

--- Comment #13 from Steve McIntyre <steve.mcintyre at linaro dot org> 2013-01-10 13:49:35 UTC ---
(In reply to comment #12)
> (In reply to comment #10)
> > Hi,
> > 
> > Could you show me all the output from ldconfig -p please? I'd like to see what
> > the rest of your libs look like. Also: I'm curious why you've got something
> > under /opt - is that something you've built as part of your system, or is
> > /opt/vc a precompiled binary from somewhere else?
> > 
> > readelf -A /lib/libc.so.6
> > readelf -A /opt/vc/lib/libvcos.so
> > 
> > might also be instructive here.
> 
> /opt/vc is the common location for the userland libraries on the Raspberry Pi
> computer.
> I built both the vcgencmd binrary and the libraries it links to myself,
> natively on the same system.
> The same problem remains if I move the libraries into /usr/local/lib for
> example, any non-default library location shows the same strangeness.

OK, fine - just wanted to check!

> Here is the full output of ldconfig -p:
> 
> [root@0] [/] ldconfig -p
> 159 libs found in cache `/etc/ld.so.cache'
>     libz.so.1 (libc6) => /lib/libz.so.1
>     libz.so (libc6) => /usr/lib/libz.so
>     libvcos.so (libc6) => /opt/vc/lib/libvcos.so
>     libvchiq_arm.so (libc6) => /opt/vc/lib/libvchiq_arm.so

<snip>

>     libcidn.so (libc6, OS ABI: Linux 2.6.25) => /usr/lib/libcidn.so
>     libc.so.6 (libc6, OS ABI: Linux 2.6.25) => /lib/libc.so.6
>     libbz2.so.1.0 (libc6) => /lib/libbz2.so.1.0

<snip>

OK, so all of your libs look consistent. That's good.

> Here's the output of readelf -A /lib/libc.so.6:
> 
> [root@0] [/] readelf -A /lib/libc.so.6
> Attribute Section: aeabi
> File Attributes
>   Tag_CPU_name: "6ZK"
>   Tag_CPU_arch: v6KZ
>   Tag_ARM_ISA_use: Yes
>   Tag_THUMB_ISA_use: Thumb-1
>   Tag_FP_arch: VFPv2
>   Tag_ABI_PCS_wchar_t: 4
>   Tag_ABI_FP_rounding: Needed
>   Tag_ABI_FP_denormal: Needed
>   Tag_ABI_FP_exceptions: Needed
>   Tag_ABI_FP_number_model: IEEE 754
>   Tag_ABI_align_needed: 8-byte
>   Tag_ABI_align_preserved: 8-byte, except leaf SP
>   Tag_ABI_enum_size: int
>   Tag_ABI_HardFP_use: SP and DP
>   Tag_ABI_VFP_args: VFP registers
>   Tag_CPU_unaligned_access: v6
>   Tag_Virtualization_use: TrustZone
> 
> Here's the output of readelf -A /opt/vc/lib/libvcos.so:
> 
> [root@0] [/] readelf -A /opt/vc/lib/libvcos.so
> Attribute Section: aeabi
> File Attributes
>   Tag_CPU_name: "6ZK"
>   Tag_CPU_arch: v6KZ
>   Tag_ARM_ISA_use: Yes
>   Tag_THUMB_ISA_use: Thumb-1
>   Tag_FP_arch: VFPv2
>   Tag_ABI_PCS_wchar_t: 4
>   Tag_ABI_FP_denormal: Needed
>   Tag_ABI_FP_exceptions: Needed
>   Tag_ABI_FP_number_model: IEEE 754
>   Tag_ABI_align_needed: 8-byte
>   Tag_ABI_align_preserved: 8-byte, except leaf SP
>   Tag_ABI_enum_size: int
>   Tag_ABI_HardFP_use: SP and DP
>   Tag_ABI_VFP_args: VFP registers
>   Tag_CPU_unaligned_access: v6
>   Tag_Virtualization_use: TrustZone

However, I'm a little worried by the fact that all of your libs are
showing up as using the soft-float ABI (lack of ",hard-float" in the
ldconfig output), but readelf shows that these two libraries are both
meant to be HFABI (Tag_ABI_VFP_args: VFP registers). That's more
expected on a system with older binaries, but with new binutils and
new glibc all the HFABI flags should be there. However, even given
that I don't see why on earth that would be affecting the library
search *paths* in the way you're seeing. If you had a mix of ABIs in
your libraries, that would be blocking incompatible loading
etc. regardless of where they live.

I'm going to do an LFS build on a machine locally now so that I can
see what on earth is going on, but that's going to take a short
while. At least I have a reasonably fast v7 machine to work with
here...

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.