Analyze¶

The analyze command is a front-end to various tools that let us analyze package installations. Each analyzer is a module for a different kind of analysis that can be done on a package installation, including (but not limited to) binary, log, or text analysis. Thus, the analyze command group allows you to take an existing package install, choose an analyzer, and extract some output for the package using it.

Analyzer Metadata¶

For all analyzers, we write to an analyzers folder in ~/.spack, or the value that you specify in your spack config at config:analyzers_dir. For example, here we see the results of running an analysis on zlib:

$ tree ~/.spack/analyzers/
└── linux-ubuntu20.04-skylake
    └── gcc-9.3.0
        └── zlib-1.2.11-sl7m27mzkbejtkrajigj3a3m37ygv4u2
            ├── environment_variables
            │   └── spack-analyzer-environment-variables.json
            ├── install_files
            │   └── spack-analyzer-install-files.json
            └── libabigail
                └── spack-analyzer-libabigail-libz.so.1.2.11.xml

This means that you can always find analyzer output in this folder, and it is organized with the same logic as the package install it was run for. If you want to customize this top level folder, simply provide the --path argument to spack analyze run. The nested organization will be maintained within your custom root.

Listing Analyzers¶

If you aren’t familiar with Spack’s analyzers, you can quickly list those that are available:

$ spack analyze list-analyzers
install_files            : install file listing read from install_manifest.json
environment_variables    : environment variables parsed from spack-build-env.txt
config_args              : config args loaded from spack-configure-args.txt
libabigail               : Application Binary Interface (ABI) features for objects

In the above, the first three are fairly simple - parsing metadata files from a package install directory to save

Analyzing a Package¶

The analyze command, akin to install, will accept a package spec to perform an analysis for. The package must be installed. Let’s walk through an example with zlib. We first ask to analyze it. However, since we have more than one install, we are asked to disambiguate:

$ spack analyze run zlib
==> Error: zlib matches multiple packages.
  Matching packages:
    fz2bs56 zlib@1.2.11%gcc@7.5.0 arch=linux-ubuntu18.04-skylake
    sl7m27m zlib@1.2.11%gcc@9.3.0 arch=linux-ubuntu20.04-skylake
  Use a more specific spec.

We can then specify the spec version that we want to analyze:

$ spack analyze run zlib/fz2bs56

If you don’t provide any specific analyzer names, by default all analyzers (shown in the list-analyzers subcommand list) will be run. If an analyzer does not have any result, it will be skipped. For example, here is a result running for zlib:

$ ls ~/.spack/analyzers/linux-ubuntu20.04-skylake/gcc-9.3.0/zlib-1.2.11-sl7m27mzkbejtkrajigj3a3m37ygv4u2/
spack-analyzer-environment-variables.json
spack-analyzer-install-files.json
spack-analyzer-libabigail-libz.so.1.2.11.xml

If you want to run a specific analyzer, ask for it with –analyzer. Here we run spack analyze on libabigail (already installed) _using_ libabigail1

$ spack analyze run --analyzer abigail libabigail

Monitoring An Analysis¶

For any kind of analysis, you can use a spack monitor “Spackmon” as a server to upload the same run metadata to. You can follow the instructions in the spack monitor documentation to first create a server along with a username and token for yourself. You can then use this guide to interact with the server.

You should first export our spack monitor token and username to the environment:

$ export SPACKMON_TOKEN=50445263afd8f67e59bd79bff597836ee6c05438
$ export SPACKMON_USER=spacky

By default, the host for your server is expected to be at http://127.0.0.1 with a prefix of ms1, and if this is the case, you can simply add the --monitor flag to the install command:

$ spack analyze run --monitor wget

If you need to customize the host or the prefix, you can do that as well:

$ spack analyze run --monitor --monitor-prefix monitor --monitor-host https://monitor-service.io wget

If your server doesn’t have authentication, you can skip it:

$ spack analyze run --monitor --monitor-disable-auth wget

Regardless of your choice, when you run analyze on an installed package (whether it was installed with --monitor or not, you’ll see the results generating as they did before, and a message that the monitor server was pinged:

$ spack analyze --monitor wget
...
==> Sending result for wget bin/wget to monitor.