Time [in minutes]: 100 (not, that's not a typo, it took an hour and 40 minutes) Platform [OS, libraries, etc.]: Used a bare metal 'dl360' PC (Dell PowerEdge 360) from aptlab.net running Ubuntu 12.04 LTS and kernel 3.2.0 . Skill level (at least the following; tell us if you have a paper-specific skill): - I can build complex software like GCC and the Linux kernel Sequence of steps to build: There were perhaps more than a hundred steps involved in the building of this software, so I'm not documenting all of them here. I followed the instructions provided at: http://research.cs.wisc.edu/sonar/projects/symdrive/downloads.shtml I have a typescript file for the process, but it is over 600K I did all steps up through and including the "Getting the Disk Image" section; I did not attempt to actually test any drivers, as the scope of this study is simply to build the software. Detailed evaluation: tl;dr: Compiling this software is possible, but difficult and time consuming. Compiling and running this software is definitely not for the faint of heart, however, in the end, it seemed to work for me. (Though, I did not attempt to actually test the software itself once all of the build and setup were completed.) It took 21 GB of disk space. This software has a number of gigantic dependencies (including s2e, qemu, a specific version of Linux kernel, cil, and the entire Debian base system) that are complicated to compile, and take a very long time and require a very large amount of disk space. However, the process is documented in detail on the paper website, so someone with sufficient motivation and experience building large, complicated codebases should be able to do so. The process could probably be improved by some scripting and packaging on the part of the software's authors. There are also places in the documentation where there is ambiguity about what directory commands should be run in, whether one should be in the chroot environment that the software creates, etc. However, I did not have great difficulty in figuring these steps out. There were a few places where peculiarities of my environment (eg. the location of my home directory) didn't match assumptions in the instructions or code. These were pretty easy to fix with minimal sleuthing. I did find myself having to go back in the process a couple of times and repeating an earlier step when it turned out I had missed something or made a mistake. The large number of dependencies do make the process seem somewhat fragile, and vulnerable to the dependencies disappearing from the net. In order to make this software more likely to still be runnable in the future, it would be highly desirable to package it all up and provide archives of all the dependencies. I can definitely understand why Collberg et al. marked this package as unbuildable given their methodology of strictly limiting build time. A package this complex cannot be built within half an hour. It also takes someone who is quite determined, at least somewhat familiar with building large systems, and willing to troubleshoot numerous small oddities in the build system(s). It's definitely not something that someone without a real interest in repeating / building on this research should undertake.