This document was last edited Feb-8-2024
daq-buildtools
is the toolset to simplify the development of DUNE DAQ packages. It provides environment and building utilities for the DAQ Suite.
To get set up, you'll need access to the cvmfs areas /cvmfs/dunedaq.opensciencegrid.org
and /cvmfs/dunedaq-development.opensciencegrid.org
. This is the case, e.g., on the np04 cluster at CERN. By default, the environment variable SPACK_DISABLE_LOCAL_CONFIG
is set to true
, which allows spack to skip user configurations found under ~/.spack
. In rare cases, if you want spack to pick up your local configurations, you can unset this environment variable by issuing unset SPACK_DISABLE_LOCAL_CONFIG
. However, the recommended way of using customized spack configrations is via the use of a workarea with a local spack instance. Customizations to spack configurations can be directly put into that local instance. More details about how to create a workarea with a spack instance can be found in the "[Advanced dbt-create options section](#Advanced dbt-create options)" on this page.
daq-buildtools
Simply do:
source /cvmfs/dunedaq.opensciencegrid.org/setup_dunedaq.sh
setup_dbt latest
Here v7.6.0 is the latest daq-buildtools version as of Feb-8-2024.
After running these two commands, then you'll see something like:
Added /cvmfs/dunedaq.opensciencegrid.org/tools/dbt/v7.6.0/bin -> PATH
Added /cvmfs/dunedaq.opensciencegrid.org/tools/dbt/v7.6.0/scripts -> PATH
DBT setuptools loaded
If you type dbt-
followed by the <tab>
key you'll see a listing of available commands, which include dbt-create
, dbt-build
, dbt-setup-release
and dbt-workarea-env
. These are all described in the following sections.
Each time that you log into a fresh Linux shell and want to either (1) set up an existing cvmfs-based DUNE DAQ software release or (2) develop code within a pre-existing DUNE DAQ work area, you'll need to set up daq-buildtools. These two cases are described in detail momentarily. For (1) you'd want to repeat the method above to set up daq-buildtools. For (2) it's easier instead to cd
into the work area and source the file named env.sh
.
If you only want access to a DUNE DAQ software release (its executables, etc.) without actually developing DUNE DAQ software itself, you'll want to run a release from cvmfs. Please note that in general, frozen releases (especially patch frozen releases) are intended for this scenario, and not for development. After setting up daq-buildtools, you can simply run the following command if you wish to use a frozen release:
dbt-setup-release <release> # fddaq-v4.2.1-a9 the latest FD frozen release as of Nov-30-2023
As of July 2023, the DUNE DAQ software stack has been split into far detector and near detector-specific parts. Starting with the v4.1.0
release of the stack, do not use the traditional convention of dunedaq-vX.Y.Z
as the frozen release label, but instead, fddaq-vX.Y.Z
and nddaq-vX.Y.Z
. Note that for v4.1.0
and v4.1.1
we only have a far detector-specific stack.
Instead of a frozen release you can also set up nightly releases or candidate releases using the same arguments as are described later for dbt-create
; e.g. if you want to set up candidate release fddaq-v4.3.0-rc2-a9
you can do:
dbt-setup-release -b candidate fddaq-v4.3.0-rc2-a9
dbt-setup-release
will set up both the external packages and DAQ packages, as well as activate the Python virtual environment. Note that the Python virtual environment activated here is read-only.
If you wish to develop DUNE DAQ software, you can start by creating a work area. Find a directory in which you want your work area to be a subdirectory (home directories are a popular choice) and cd
into that directory. Then think of a good name for the work area (give it any name, but we'll refer to it as "MyTopDir" in this document).
Each work area is based on a DUNE DAQ software release, which defines what external and DUNE DAQ packages the code you develop in a work area are built against. Releases come in four categories:
develop
and production/v4
branches. Depending on whether it's the far detector stack or the near detector stack, and whether it's a develop or production build, these are generally labeled either as NFD_<branch>_<YY><MM><DD>_<OS>
(far detector) or NND_<branch>_<YY><MM><DD>_<OS>
(near detector). E.g. NFD_DEV_240213_A9
is the AL9 nightly develop build for the far detector on February 13th, 2024, while NFD_PROD4_240213_C8
is the CentOS nightly production build for the far detector on the same date. Note that, prior to February 13th, 2024, the labeling convention is NFD<YY>-<MM>-<DD>
for the far detector stack and NND<YY>-<MM>-<DD>
for the near detector stack. fddaq-vX.Y.X
or nddaq-vX.Y.Z
. fddaq-vX.Y.Z-rc<candidate iteration>-<OS>
or nddaq-vX.Y.Z-rc<candidate iteration>-<OS>
. For example, fddaq-v4.3.0-rc2-a9
is the second release candidate for the AL9 build of fddaq-v4.3.0
. Note that, prior to February 13th, 2024, the labeling convention is fddaq-vX.Y.Z-c<candidate iteration>
The majority of work areas are set up to build against the most recent nightly release. To do so, run:
dbt-create -n <nightly release> <name of work area subdirectory> # E.g., NFD_DEV_240213_A9
You can also use -n last_fddaq
or -n last_nddaq
to build against the most recent production branch, e.g., NFD_PROD4_240213_A9
. To see all available nightly releases, run dbt-create -l -n
or dbt-create -l -b nightly
. Note also that you can leave out defining the name of the work area subdirectory, in which case it defaults to the same name as the release.
If you want to build against a candidate release, run:
dbt-create -b candidate <candidate release> <name of work area subdirectory> # E.g., fddaq-v4.3.0-rc2-a9
...where to see all available candidate releases, run dbt-create -l -b candidate
.
And to build against a frozen release (not recommended, as the codebase changes fairly rapidly), you don't need the -b <release type>
argument at all. You can just do:
dbt-create <frozen release> <name of work area subdirectory>
The structure of your work area will look like the following:
MyTopDir
├── build
├── dbt-workarea-constants.sh
├── env.sh
├── log
└── sourcecode
├── CMakeLists.txt
└── dbt-build-order.cmake
The next section of this document concerns how to build code in your new work area. However, if you'd like to learn about how to retrieve information about your work area such as the release of the DUNE DAQ suite it builds against, you can skip ahead to Finding Info on Your Work Area.
dbt-create
optionsAlong with telling dbt-create
what you want your work area to be named and what release you want it to be based off of, there are a few more options that give you finer-grained control over the work area. You can simply run dbt-create -h
for a summary, but they're described in fuller detail here.
-s/--spack
: Install a local Spack instance in the work area. This will allow you to install and load whatever Spack packages you wish into your work area.
-q/--quick
: Use this if you don't plan to develop a Python package. This is much quicker than the default behavior of dbt-create, which will actually copy the Python virtual environment over to your work area, thereby giving you write permission to the project's Python packages. With -q/--quick
, the Python virtual environment your work area uses is in the (read-only) release area on cvmfs.
-i/--install-pyvenv
: With this option, there will be compilation/installation of python modules using the pyvenv_requirements.txt
in the release directory. This is typically slower than cloning, but not always. You can take further control by combining it with the -p <requirements file>
argument, though it's unlikely as a typical developer that you'd want a non-standard set of Python packages.
First step: cd
into the base of the work area you've created.
For the purposes of instruction, let's build the listrev
package. Downloading it is simple:
cd sourcecode
git clone https://github.com/DUNE-DAQ/listrev.git
cd ..
Note that in a "real world" situation you'd be doing your development on a feature branch in which case you'd add -b <branch you want to work on>
to the git clone
command above.
We're about to build and install the listrev
package. (🔴 Note: if you are working with other packages, have a look at the Working with more repos subsection before running the following build command.) By default, the scripts will create a subdirectory of MyTopDir called ./install
and install any packages you build off your repos there. If you wish to install them in another location, you'll want to set the environment variable DBT_INSTALL_DIR
to the desired installation path before source-ing the env.sh
script described below. You'll also want to remember to set the variable during subsequent logins to the work area if you don't go with the default.
Now, do the following:
. ./env.sh # To set up your work area's environment
dbt-build
...and this will build listrev
in the local ./build
subdirectory and then install it as a package either in the local ./install
subdirectory or in whatever you pointed DBT_INSTALL_DIR
to. env.sh
performs two steps: it will both set up the daq-buildtools environment (if it hasn't already been set) and then it will update environment variables (LD_LIBRARY_PATH
, etc.) to account for the packages in your work area. Note that whenever you add a new repo to your work area, you'll want to run the second of these two steps, dbt-workarea-env
, so that environment variables such as LD_LIBRARY_PATH
, etc, are again updated accordingly.
To work with more repos, add them to the ./sourcecode
subdirectory as we did with listrev. Be aware, though: if you're developing a new repo which itself depends on another new repo, daq-buildtools may not already know about this dependency. "New" in this context means "not listed in /cvmfs/dunedaq.opensciencegrid.org/spack/releases/fddaq-v4.1.1/dbt-build-order.cmake
". If this is the case, add the names of your new package(s) to the build_order
list found in ./sourcecode/dbt-build-order.cmake
, placing them in the list in the relative order in which you want them to be built.
As a reminder, once you've added your repos and built them, you'll want to run dbt-workarea-env
so the environment picks up their applications, libraries, etc.
dbt-build
will by default skip CMake's config+generate stages and go straight to the build stage unless either the CMakeCache.txt
file isn't found in ./build
or you've just added a new repo to ./sourcecode
. If you want to remove all the contents of ./build
and run config+generate+build, all you need to do is add the --clean
option, i.e.
dbt-build --clean
One case where you'd want to do this is if you changed the installation directory variable as described above.
And if, after the build, you want to run the unit tests, just add the --unittest
option. Note that it can be used with or without --clean
, so, e.g.:
dbt-build --clean --unittest # Blow away the contents of ./build, run config+generate+build, and then run the unit tests
..where in the above case, you blow away the contents of ./build
, run config+generate+build, install the result in $DBT_INSTALL_DIR
and then run the unit tests. Be aware that for many packages, unit tests will only (fully) work if you've also rerun dbt-workarea-env
.
To check for deviations from the coding rules described in the DUNE C++ Style Guide, run with the --lint
option:
dbt-build --lint
...though be aware that some guideline violations (e.g., having a function which tries to do unrelated things) can't be picked up by the automated linter. Also note that you can use dbt-clang-format.sh
in order to automatically fix whitespace issues in your code; type it at the command line without arguments to learn how to use it.
Note that unlike the other options to dbt-build
, --lint
and --unittest
are both capable of taking an optional argument, which is the name of a specific repo in your work area which you'd like to either lint or run unit tests for. This can be useful if you're focusing on developing one of several repos in your work area; e.g. dbt-build --lint <repo you're working on>
. With --lint
you can get even more fine grained by passing it the name of a single file in your repository area; either the absolute path for the file or its path relative to the directory you ran dbt-build
from will work.
If you want to see verbose output from the compiler, all you need to do is add the --cpp-verbose
option:
dbt-build --cpp-verbose
If you want to change cmake message log level, you can use the --cmake-msg-lvl
option:
dbt-build --cmake-msg-lvl=<ERROR|WARNING|NOTICE|STATUS|VERBOSE|DEBUG|TRACE>
By default the build is performed using gcc's O2
compilation flag. If you wish to use a different flag, you can use the --optimize-flag
argument, e.g.:
dbt-build --optimize-flag O3 # Or Og, etc.
If you wish to only generate files but not also perform a compilation (this is a kind of expert action, but there are use cases for it) you can run:
dbt-build --codegen-only
You can see all the options listed if you run the script with the --help
command, i.e.
dbt-build --help
Finally, note that both the output of your builds and your unit tests are logged to files in the ./log
subdirectory. These files will have ASCII color codes which make them difficult to read with some tools; less -R <logfilename>
, however, will display the colors and not the codes themselves.
In order to access the applications, libraries and plugins built and installed into the $DBT_INSTALL_DIR
area during the above procedure, the system needs to be instructed on where to look for them. This is accomplished via tha env.sh
file you've already seen. E.g., log into a new shell, cd into your work area, then do the following:
export DBT_INSTALL_DIR=<your installation directory> # ONLY needed if you didn't use the default
. ./env.sh
Note that if you add a new repo to your work area, after building your new code - and hence putting its output in ./build
- you'll need to run dbt-workarea-env
.
Once the runtime environment is set, just run the application you need. listrev, however, has no applications; it's just a set of DAQ module plugins which get added to CET_PLUGIN_PATH.
Now that you know how to set up a work area, a classic option for learning about how to run DAQ modules in a work area is the listrev documentation.
In the listrev instructions you'll be running a program called nanorc
to run the DAQ. To learn more about nanorc
itself, take a look at the nanorc documentation.
A couple of things need to be kept in mind when you're building code in a work area. The first is that when you call dbt-build
, it will build your repos against a specific release of the DUNE DAQ software stack - namely, the release you (or someone else) provided to dbt-create
when the work area was first created. Another is that the layout and behavior of a work area is a function of the version of daq-buildtools which was used to create it. As a work area ages it becomes increasingly likely that a problem will occur when you try to build a repo in it; this is natural and unavoidable.
As such, it's important to know the assumptions a work area makes when you use it to build code. This section covers ways to learn details about your work area and its contents.
dbt-info
A useful script to call to get immediate information on your development environment is dbt-info
. For a full set of options you can simply run dbt-info --help
, but for a quick summary, we have the following:
dbt-info release
: tells you if it's a far detector or near detector release, what its name is (e.g. NFD_DEV_240213_A9
), what the name of the base release is, and where the release is located in cvmfs.
dbt-info package <package name>
: tells you info about the DUNE DAQ package whose name you provide it (git commit hash of its code, etc.). Passing "all" as the package name gives you info for all the DUNE DAQ packages.
dbt-info external <package name>
: external
is same as the package
option, except you use it when you want info not on a DUNE DAQ package but an external package (e.g., boost
)
dbt-info pymodule <python module>
: get the version of a Python module. Response will differ depending on whether you have a local Python environment in your work area.
dbt-info sourcecode
: will tell you the branch each of the repos in your work area is on, as well as whether the code on the branch has been edited (indicated by an *
)
dbt-info release_size
: tells you the # of packages and memory (in KB) used by each of the release, the base release, and the externals.
dbt-workarea-constants.sh
In the base of your work area is a file called dbt-workarea-constants.sh
, which will look something like the following:
export SPACK_RELEASE="fddaq-v4.1.0"
export SPACK_RELEASES_DIR="/cvmfs/dunedaq.opensciencegrid.org/spack/releases"
export DBT_ROOT_WHEN_CREATED="/cvmfs/dunedaq.opensciencegrid.org/tools/dbt/v7.2.1"
export LOCAL_SPACK_DIR="/home/jcfree/daqbuild_fddaq-v4.1.0/.spack"
This file is sourced whenever you run dbt-workarea-env
, and it tells both the build system and the developer where they can find crucial information about the work areas' builds. Specifically, these environment variables mean the following:
$SPACK_RELEASE
: this is the release of the DUNE DAQ software stack against which repos will build (e.g. fddaq-v4.3.0-rc2-a9
, NFD_DEV_240213_A9
, etc.)$SPACK_RELEASES_DIR
: The base of the directory containing the DUNE DAQ software installations. DBT_ROOT_WHEN_CREATED
: The directory containing the env.sh
file which was sourced before this work area was first createdLOCAL_SPACK_DIR
: If the -s/--spack
was passed to dbt-create
when the work area was built, this points to where the local Spack area is locatedThere are also useful Spack commands which can be executed to learn about the versions of the individual packages you're working with, once you've run dbt-workarea-env
or dbt-setup-release
. An excellent Spack tutorial inside the official Spack documentation is worth a look, but a few Spack commands can be used right away to learn more about your environment. They're presented both for the case of you having set up a nightly release and a frozen release:
spack find --loaded -N | grep coredaq-vX.Y.Z
or spack find --loaded -N | grep NB
will tell you all the DUNE DAQ packages shared by both far- and near detector software which have been loaded by dbt-workarea-env
or dbt-setup-release
spack find --loaded -N | grep fddaq-vX.Y.Z
or spack find --loaded -N | grep FD
for far detector DUNE DAQ packagesspack find --loaded -N | grep nddaq-vX.Y.Z
or spack find --loaded -N | grep ND
for near detector DUNE DAQ packagesspack find --loaded -N | grep dunedaq-externals
for external packages not developed by DUNE collaboratorsspack find --loaded -p <package name>
will tell you the path to the actual contents of a Spack-installed packageFinally, when dbt-build
is run, a file called daq_app_rte.sh
is
produced and placed in your installation area ($DBT_INSTALL_DIR
). You generally don't need to think about daq_app_rte.sh
unless you're curious; it's a sourceable file which contains environment variables that nanorc uses to launch processes when performing runs.