-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 This is a regression of http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=531522
The checks for FAKEROOT introduced to fix this bug upstream are preceeded by a new check that calls stat re-producing the original bug. Re-ordering the checks to ensure that FAKEROOT is detected first appears to resolve this issue - please see the attached patch. - -- James Page Ubuntu Core Developer -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQIcBAEBCAAGBQJPO8IMAAoJEL/srsug59jDA9EP/21MdxOTHZId7Iu6Luuhs0HD VSWIVcS8I+XmJdfuZLyGvjt7dsjezqAcJKU9wHmvPl/b3EVlBYkgSCwmdzZRp/0Y 7YJOyxcCMFyhuQxV6uYMVMyMTUzADHFAB64005ksnL+1I+Wue7hEJFTRsf1KZBQD osME0nAh5wsaEMGwL85ucVkUm1i7030sey1aSE3/WP0ZxGZ90tCkxnobTqNSlC7z /M7zul2qjVtn/M0kHwqjTxyNDc/CzNrT38zvyKttPKp49bfz3yWWVt6n2Bbna6y5 PffVA8bpldFNtMYZJn0ORLMp0X59XSlpYxw/7reB7wSXvOcHjAZP7F+t2f35ODFZ nE6xhhsNEHBg4Q7RzTzShFGHmVTBPDm1u27aznwY48VLkFIQnalXReyRqyr49E8E 6zk07dWCvdYhvTTlz+pPs2e45eYK7qa15TM6qzkdAswpK+d2d30EMR/NoaCr7yuE M4dw/6zZvRDTg8f23YPGqVvlGEMAfaHQe5Hf4S4wSgj3OlQslDk7E03RXjtSj4dp 9DbPtBJjn6slKlWfqFh0+olrD3uPqTa8H3TqWSe34+LZvFCvCLQY/Q17EtM6XUQC mJ13yJsntwvp/tOblxoFezeX4w8J5M0DQgdJTm/eJQl602Y+tcVhHvlp6WgQN6ok Dfxvx5PgSyc85KmEooWB =vMQs -----END PGP SIGNATURE-----
Description: If running under Debian fakeroot then its important that this hook is disabled; a new stat check was introduced in 1.5.x which preceeds the FAKEROOT check resulting in a SEGFAULT. Author: James Page <james.p...@ubuntu.com> Origin: https://svn.open-mpi.org/trac/ompi/changeset/21489 Index: openmpi-1.5.4/opal/mca/memory/linux/hooks.c =================================================================== --- openmpi-1.5.4.orig/opal/mca/memory/linux/hooks.c 2012-02-15 12:59:39.986314651 +0000 +++ openmpi-1.5.4/opal/mca/memory/linux/hooks.c 2012-02-15 13:14:57.744722517 +0000 @@ -738,15 +738,6 @@ check_result_t r1, r2, lp, lpp; bool want_rcache = false, found_driver = false; - /* First, check if ummunotify is present on the system. If it is, - then we don't need to do the following ptmalloc2 hacks. - open/mmap on the device may fail during init, but if /dev/ummunotify - exists, we assume that the user/administrator *wants* to use - ummunotify. */ - if (stat("/dev/ummunotify", &st) == 0) { - return; - } - /* Yes, checking for an MPI MCA parameter here is an abstraction violation. Cope. Yes, even checking for *any* MCA parameter here (without going through the MCA param API) is an @@ -769,6 +760,15 @@ return; } + /* Next, check if ummunotify is present on the system. If it is, + then we don't need to do the following ptmalloc2 hacks. + open/mmap on the device may fail during init, but if /dev/ummunotify + exists, we assume that the user/administrator *wants* to use + ummunotify. */ + if (stat("/dev/ummunotify", &st) == 0) { + return; + } + /* Look for sentinel files (directories) to see if various network drivers are loaded (yes, I know, further abstraction violations...).