Re: [commit] [patch] Fix dwfl_report_elf BASE alignment [Re: pending patches ping#2]

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Wed, 20 Mar 2013 16:31:41 +0100, Jan Kratochvil wrote:
> One needs to
> disable build-ids from the core file.  I tried it the official way of bit 4
> in coredump_filter but this does not work, filed kernel Bug:
> 	https://bugzilla.redhat.com/show_bug.cgi?id=923827

This is no bug, I just used (instead of default 0x33) wrong mask, the right
one should be apparently 0x23.


But I found I have regressed elfutils by:
	[commit] [patch] Fix for PIE with both -e and --core in use
	https://lists.fedorahosted.org/pipermail/elfutils-devel/2012-November/002793.html
	commit d5784afaf49cd59b9286e766cd9b1cf00cb43553

before:
$ ./src/addr2line -S -e /bin/sleep --core=/tmp/sleep.core.23 0x00007fefc0c2b430
__gconv_lookup_cache+0xb0
/usr/src/debug/glibc-2.16-75f0d304/iconv/gconv_cache.c:163
 - although the symbol is wrong due to this bug

and although still before it required proper args order:
./src/addr2line -S --core=/tmp/sleep.core.23 0x00007fefc0c2b430 -e /bin/sleep
??
??:0

but now always:
??
??:0

because:
libdwfl/link_map.c
            /* We could not read the executable's phdrs from the
               memory image.  If we have a presupplied executable,
               we can still use the AT_PHDR and AT_ENTRY values to
               verify it, and to adjust its bias if it's a PIE.

There is no executable to check available yet after the patch above.


To be fixed later.


Jan