Crashes when installing on Ubuntu 16.10

[Vitor Gouveia]

When I try to install gridengine-master on Ubuntu 16.10 I got the following error:

$ sudo apt-get install gridengine-client gridengine-master
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following additional packages will be installed:
  db-util gridengine-common libhwloc-plugins libhwloc5
Suggested packages:
  gridengine-qmon libhwloc-contrib-plugins
The following NEW packages will be installed
  db-util gridengine-client gridengine-common gridengine-master libhwloc-plugins libhwloc5
0 to upgrade, 6 to newly install, 0 to remove and 0 not to upgrade.
Need to get 0 B/8,902 kB of archives.
After this operation, 51.2 MB of additional disk space will be used.
Do you want to continue? [Y/n] y
Preconfiguring packages ...
Selecting previously unselected package libhwloc5:amd64.
(Reading database ... 409974 files and directories currently installed.)
Preparing to unpack .../0-libhwloc5_1.11.3-2_amd64.deb ...
Unpacking libhwloc5:amd64 (1.11.3-2) ...
Selecting previously unselected package gridengine-common.
Preparing to unpack .../1-gridengine-common_8.1.9+dfsg-2_all.deb ...
Unpacking gridengine-common (8.1.9+dfsg-2) ...
Selecting previously unselected package gridengine-client.
Preparing to unpack .../2-gridengine-client_8.1.9+dfsg-2_amd64.deb ...
Unpacking gridengine-client (8.1.9+dfsg-2) ...
Selecting previously unselected package gridengine-master.
Preparing to unpack .../3-gridengine-master_8.1.9+dfsg-2_amd64.deb ...
Unpacking gridengine-master (8.1.9+dfsg-2) ...
Selecting previously unselected package db-util.
Preparing to unpack .../4-db-util_1%3a5.3.21~exp1ubuntu2_all.deb ...
Unpacking db-util (1:5.3.21~exp1ubuntu2) ...
Selecting previously unselected package libhwloc-plugins.
Preparing to unpack .../5-libhwloc-plugins_1.11.3-2_amd64.deb ...
Unpacking libhwloc-plugins (1.11.3-2) ...
Processing triggers for ureadahead (0.100.0-19) ...
Setting up libhwloc5:amd64 (1.11.3-2) ...
Setting up db-util (1:5.3.21~exp1ubuntu2) ...
Setting up libhwloc-plugins (1.11.3-2) ...
Processing triggers for libc-bin (2.24-3ubuntu1) ...
Processing triggers for systemd (231-9git1) ...
Processing triggers for man-db (2.7.5-1) ...
Setting up gridengine-common (8.1.9+dfsg-2) ...

Creating config file /etc/default/gridengine with new version
Setting up gridengine-client (8.1.9+dfsg-2) ...
Setting up gridengine-master (8.1.9+dfsg-2) ...
Initializing cluster with the following parameters:
 => SGE_ROOT: /var/lib/gridengine
 => SGE_CELL: default
 => Spool directory: /var/spool/gridengine/spooldb
 => Initial manager user: sgeadmin
Initializing spool (/var/spool/gridengine/spooldb)
Initializing global configuration based on /usr/share/gridengine/default-configuration
Segmentation fault (core dumped)
dpkg: error processing package gridengine-master (--configure):
 subprocess installed post-installation script returned error exit status 139
Processing triggers for systemd (231-9git1) ...
Processing triggers for ureadahead (0.100.0-19) ...
Errors were encountered while processing:
 gridengine-master
E: Sub-process /usr/bin/dpkg returned an error code (1)

I could successfully install the same package coming from Ubuntu 16.04

[Afif Elghraoui (afif-r)]

Hmm. According to your output, the command that segfaulted was:

/usr/lib/gridengine/spooldefaults configuration /usr/share/gridengine/default-configuration

I've directly tested the installation of this package on Debian stable-backports, and Debian QA tools report successful installation/uninstallation on Debian Unstable and Testing. I suspect it might either have to do with the versions of some dependencies that were incorporated into Ubuntu 16.10, or perhaps some unexpected data in your configuration. I don't know for sure.

When you said you could successfully install on 16.04, were you referring to the version of the package distributed there, or did you try the same version as here, 8.1.9?

[Vitor Gouveia (vitorgg)]

Sorry if I wasn't clear.
The package I was talking about is 6.2u5-7.4 currently available in the repositories of Ubuntu 16.04.
I cannot perform a clean installation of gridengine-client gridengine-master (8.1.9+dfsg-2 available) on Ubuntu 16.10 but I cannot perform a clean installation of the packages coming from the Ubuntu 16.04 (gridengine-client gridengine-master 6.2u5-7.4).
I didn't try with debian but something is wrong with package or with dependencies. I'm trying to perform a clean installation.

[Gilles Degottex (gilles-degottex)]

Same problem here with Ubuntu 17.04 server.
Any news since then?

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Gilles Degottex (gilles-degottex)]

For information:
master and execution host binaries can be compiled from the source (gridengine_8.1.9+dfsg.orig.tar.gz), installed, and running flawlessly.

deb pkg gridengine-client can then be used on submit hosts.

Thus, the bug is surely from the configuration script of the pkg gridengine-master

[Marco van Zwetselaar (zwets)]

Problem still exists with Ubuntu 17.10 server. Segfault during standard install of gridengine-master.

[Ewald Enzinger (entn-at-deactivatedaccount)]

Problem still exists with Ubuntu 18.04 LTS. Segfault during standard install of gridengine-master. Same error message.

[Ewald Enzinger (entn-at-deactivatedaccount)]

For information:
Installing gridengine-master, gridengine-common, gridengine-exec, gridengine-client version 8.1.9+dfsg-8 from Debian sid (https://packages.debian.org/sid/utils/gridengine-master) solves the problem on Ubuntu 18.04 LTS.

[Vladimir Fonov (vladimir-fonov)]

I tried poking around with the source package on a machine recently upgraded from 16.04.
I had following observations:

1. compilation from source fails, because default-jdk which is used to compile java code doesn't support -d64 switch anymore, using openjdk-8-jdk seem to have fixed it

2. linking of glob fails, because of the references to __alloca in source/3rdparty/qmake/glob/glob.c , adding # define __alloca alloca fixed it

3. running with valgrind shows some bad memory access in source/libs/uti/sge_profiling.c , fixed it by disabling profiling: "static bool profiling_enabled = false;"

4. running with gdb showed SEGFAULT in source/libs/uti/sge_hostname.c:1167 , made a work-around by replacing the line with "bool ignore_fqdn = true;"

after that newly generated package is able to install.

Somebody should investigate why calling bootstrap_get_ignore_fqdn() causes a crash.

[Lennart Karssen (l.c.karssen)]

For Ubuntu 18.04 there is a separate report for this bug: #1774302

[Mike (md97124)]

Regarding

   Installing gridengine-master, gridengine-common, gridengine-exec,
   gridengine-client version 8.1.9+dfsg-8 from Debian sid
   (https://packages.debian.org/sid/utils/gridengine-master)
   solves the problem on Ubuntu 18.04 LTS.

I cannot reproduce the solution on Ubuntu 18.04 LTS. The gridengine-client mentioned above has a dependency on libc6 >=2.28 but 2.27 is the newest available for Ubuntu 18.04 LTS.

[Christoph Probst (christophprobst)]

There is a workaround for Ubuntu 18.04 Bionic manually replacing spooldefaults.bin, libspoolb.so and libspoolc.so with the files from the Debian 8.1.9+dfsg-9 package.

For details see https://bugs.launchpad.net/ubuntu/+source/gridengine/+bug/1774302