https://sourceware.org/bugzilla/show_bug.cgi?id=32542

--- Comment #6 from Tom Tromey <tromey at sourceware dot org> ---
(In reply to Alan Modra from comment #5)
> If this is a bfd problem it ought to be reproducible with objdump

Indeed.

I ran gdb under valgrind and also set worker-threads to 0.
It crashed but there were no valgrind reports.

Looking into the core file, I found that this frame in the
stack trace:

#6  0x0000555559de857e in solib_bfd_open (pathname=0x50300005b720
"/lib/ld-musl-mips-sf.so.1") at /home/simark/src/binutils-gdb/gdb/solib.c:456

... is somewhat misleading.  This file is not in fact used verbatim:

(gdb) p *abfd.m_obj
$3 = {
  filename = 0x1b9ec3a0 "target:/lib/ld-musl-mips-sf.so.1",

That is, this is the library as sent over the remote connection.

If I do this:

(gdb) set sysroot /tmp/pentesting-GL.iNET/rom-3.216/

... then "target remote :1234" does not cause a crash.

>From this, my conclusion is:

1. There's a bug in qemu
2. There's probably still a bug in BFD, if we can dig out the contents
   of whatever is being sent from qemu.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

Reply via email to