Also running :
 gdb /lib/ld-linux-armhf.so.3
...
(gdb) run /usr/lib/libavcodec.so.58.18.100
Starting program: /usr/lib/ld-linux-armhf.so.3 /usr/lib/libavcodec.so.58.18.100

Program received signal SIGSEGV, Segmentation fault.
0xb6fd7b90 in do_lookup_x ()
(gdb) bt
#0  0xb6fd7b90 in do_lookup_x ()
#1  0xb6fd8490 in _dl_lookup_symbol_x ()
#2  0xb6fda7b8 in _dl_relocate_object ()
#3  0xb6fd1c4c in dl_main ()
#4  0xb6fe544c in _dl_sysdep_start ()
#5  0x00000064 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

You've provided no evidence that the problem is in any way related to gcc.

(In reply to Jakub Jelinek from comment #2)
> You've provided no evidence that the problem is in any way related to gcc.
And I've indeed no evidence except that this seems to be a "toolchain issue"
because the same ffmpeg package works in f30 and f28 but not on f29 for armhfp.

I'm currently trying to reproduce the issue using ffmpeg-4.0.2-1.fc29 into the buildroot (used to build the last vlc on armhfp).
Then I'm going to resubmit a scratch build of ffmpeg-4.0.2-1.fc29 to see if the same error can be seen.

I'm tryint to audit the dependencies of libavcodec and this shows:

for i in $(ldd -r -d /usr/lib/libavformat.so.58 | awk '! /ld-linux-armhf.so.3/ {print $1}') ; do  ret=0 ; ldd -d -r /usr/lib/$i &>/dev/null; ret=$?; if [ ! x"$ret" = x"0" ] ; then echo $i ; fi; done
libavcodec.so.58
libvo-amrwbenc.so.0

But then I don't see what's the issue with libvo-amrwbenc.so.0
# ldd -r -d /usr/lib/libvo-amrwbenc.so.0
	linux-vdso.so.1 (0xbec35000)
	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0xb6ef7000)
	libc.so.6 => /lib/libc.so.6 (0xb6dab000)
	/lib/ld-linux-armhf.so.3 (0xb6f5c000)
<mock-chroot> sh-4.4# echo $?
1

ldd -r -d ./usr/lib/libvo-amrwbenc.so.0 from a fedora 27 armhfp kernel+userspace return 0, whereas in the same kernel but within mock for 29, it return 1

Tried using this build:
http://koji.rpmfusion.org/kojifiles/packages/vo-amrwbenc/0.1.3/6.fc29/armv7hl/vo-amrwbenc-0.1.3-6.fc29.armv7hl.rpm

I've rebuilded vo-amrwebenc and it's worked. so the previous built broke the dependencies. I will try to reproduce without adding the older ffmpeg.

So something went wrong at some point, and I'm not sure which others build are broken on fedora side or elsewhere...

Only updating the fixed vo-amrwbenc-0.1.3-8.fc29.armv7hl package was enough to clear the rpmlint error on ffmpeg-libs (ldd-failed /usr/lib/libavcodec.so.58.18.100).

So the problem was a bad compilation of the vo-amrwbenc package (built on 2018-08-19 20:19:38 on arm-builder09.home.rpmfusion.net).
This built was part of the second mass-rebuild batch made to clear the other toolchain issue. So far, no

I haven't seen any other case of such behaviour or any others fedora/non-fedora packages for f29 armhfp (from a fedora 29 workstation image), so closing.