Re: port-evbarm/48738: Upgrade of -current kernel on Raspberry Pi brakes running executables from previous version requiring reinstall

[Martin Husemann <martin%duskware.de@localhost>]

The following reply was made to PR port-evbarm/48738; it has been noted by 
GNATS.

From: Martin Husemann <martin%duskware.de@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: port-evbarm/48738: Upgrade of -current kernel on Raspberry Pi 
brakes running executables from previous version requiring reinstall
Date: Sat, 12 Apr 2014 18:40:33 +0200

 On Sat, Apr 12, 2014 at 10:15:00AM +0000, joachim.thiemnn%gmail.com@localhost 
wrote:
 > Either upgrade all of userland, or use kernel option COMPAT_NETBSD32.
 > The latter can still cause problems though, with dynamic libraries (I
 > think).
 
 Did you try COMPAT_NETBSD32?
 
 This should only help if you changed from an old "arm" install to "earm"
 now - which might need a bit more planning and library copying (and would
 make it no bug).
 
 If you have an old binary around still, you can check the ABI by running
 file on it. On my eabrm machine for example file /bin/sh says:
 
 /bin/sh: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV), dynamically 
linked (uses shared libs), for NetBSD 6.99.40, compiled for: earmv4, not 
stripped
 
 Another possible issue are broken binaries with corrupt ELF notes (I've seen
 that on test machines I updated early to gcc 4.8 - not sure if this ever
 happend in the official builds). Symptom here would be a corrupt binary
 that old kernels erroneously accepted, but newer kernels reject. You
 can verify this by running readelf -n on the binary. Again /bin/sh as
 an example:
 
 Notes at offset 0x00000128 with length 0x00000018:
   Owner                 Data size       Description
   NetBSD        0x00000004      IDENT 699004000 (6.99.40)
 
 Notes at offset 0x00000140 with length 0x00000014:
   Owner                 Data size       Description
   PaX                  0x00000004       Unknown note type: (0x00000003)
 
 Notes at offset 0x00000154 with length 0x0000001c:
   Owner                 Data size       Description
   NetBSD        0x00000007      MARCH <earmv4>
 
 (the last one will be different on RPi).
 
 The corrupt binaries cause readelf to complain in this output.
 
 Martin