Tools Overview

Creating classes: class_maker.py

To make it faster to create a new cpp class. This is a small python script located in buildconfig/class_maker.py. It generates the .cpp, .h, and test files for a class along with some code stubs. It can also flesh out more methods for new Algorithms, using the --alg option which also creates a stub user documentation page .rst.

usage: class_maker.py [-h] [--force] [--no-header] [--no-test] [--no-cpp]
                      [--no-rst] [--no-cmake] [--alg] [--subfolder SUBFOLDER]
                      [--project PROJECT]
                      SUBPROJECT CLASSNAME

Utility to create Mantid class files: header, source and test. version 1.0

positional arguments:
  SUBPROJECT            The subproject under Framework/; e.g. Kernel
  CLASSNAME             Name of the class to create

optional arguments:
  -h, --help            show this help message and exit
  --force               Force overwriting existing files. Use with caution!
  --no-header           Don't create the header file
  --no-test             Don't create the test file
  --no-cpp              Don't create the cpp file
  --no-rst              Don't create the rst file
  --no-cmake            Don't modify cmake files
  --alg                 Create an Algorithm stub. This adds some methods
                        common to algorithms.
  --subfolder SUBFOLDER
                        Put the source under a subfolder below the main part
                        of the project, e.g. Geometry/Instrument.
  --project PROJECT     The project in which this goes. Default: Framework.
                        Can be MantidQt

Moving/Renaming classes: move_class.py

This python script is located in in /buidconfig/. It will move a class from one subproject to another and/or rename the class. Namespaces and cmakelists are adjusted. For details, run:

buildconfig/move_class.py --help

Deleting a class: delete_class.py

This python script is located in in /buildconfig/. It will delete a class from one subproject. CMakeList.txt is adjusted. For details, run:

buildconfig/delete_class.py --help

Profiling

Profiling could mean tracking start-up, the progress of an algoirthm or performance. See Profiling Overview for more details.

Leak checking etc

Linux

Memcheck

• Keeps track of allocs/deallocs and reports anything missing at exit.

• Slow but thorough

• Useful options to run with

• See valgrind for details on how to install

valgrind --tool=memcheck --leak-check=full --show-reachable=yes --num-callers=20 --track-fds=yes --track-origins=yes --freelist-vol=500000000 ``\ \ `` [args...]

Windows

Visual Leak Detector

1. Setup the additional paths as defined in the readme file

2. Adjust the configuration file, “C:Program FilesVisual Leak Detectorvld.ini” to output to both File and debugger by changing the ReportTo to

ReportTo = both

1. Add #include <vld.h> to the system.h file in Kernel

2. Compile everything in debug

3. Running unit tests should now create a file memory_leak_report.txt in the test directory.

4. IMPORTANT remove the #include <vld.ini> before checking in.

Thread checking

Helgrind or drd

• Identifies race conditions & dead-locks

• Slow but accurate

• A pain to get working with OpenMP. GCC must be recompiled to use a different call to create OMP threads or helgrind/drd cannot “see” the thread calls. Use this script to recompile the same version off gcc that is onyour system. The script will need editing to change the appropriate variables.

IWYU

include what you use (iwyu) is a clang-based tool for determining what include statements are needed in C/C++ files. Below are instructions for getting it to run with mantid on linux which is a filled in version of this bug.

1. Install the software. The version available from system installs should be fine (e.g. yum or apt-get).

2. Get a copy of iwyu_tool.py which is in the project’s repository, but may not be installed if you got it from your operating system locations (e.g. yum).

3. Run cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=TRUE. This will generate an extra file, compile_commands.json, in your build area which has instructions on compiling every file in mantid.

4. Run iwyu_tool.py -p `pwd 2> iwyu.log` to generate the report of changes redirecting into the file iwyu.log. This will take a long time since it is going through the whole repository. If you want it for a single file, then supply that as an additional argument with full path. Only one file can be supplied at a time.

5. Run fix_includes < iwyu.log and compile the results. Depending on how you installed iwyu, the program may be called fix_includes.py. If it doesn’t compile, the most likely suspect is that iwyu included a private header. See iwyu instructions for users for ways to handle this. Generally, they suggest deleting the offending lines.

6. Check that your build path didn’t make it into source files. Since compile_commands.json has full paths, iwyu will put full paths in the include statements. This will not produce an error on your system, but it will on the build servers. The easiest way to check is to use the silver searcher to check for your username in your source tree.

7. Enjoy your success.

Note: iwyu outputs to stderr and always returns a failure status code since it generates no output. The output stream also affects iwyu_tool.py

Network Related Tools

Wireshark

Linux distros should have this in their repositories. For other platforms download it from https://www.wireshark.org.

Cross-platform tool for inspecting network packets. This is useful to troubleshoot many different types of network-related problems.

Wonder Shaper

Linux only. Install it from your distro’s repository.

Wonder Shaper allows the user to limit the bandwidth of one or more network adapters. This is useful for debugging issues when a network interface is still active but very slow. More details can be found at http://xmodulo.com/limit-network-bandwidth-linux.html.

Convert Wiki Docs to Sphinx

wiki2rst reads in mediawiki formatted webpages and converts them to .rst files, for use in Sphinx. The code attempts to take all images and internal links and re-create the documentation structure in the Sphinx format.

Use

• Having added mantid to your Python path wiki2rst is run by:

python wiki2rst.py -o <output_file.rst> <url_extension>

• The <url_extension> is the part of the url after the main address (https://www.mantidproject.org/).

• There are several additional options:

  • --index_url change the name of the main url address

  • --images-dir set a relative location for the images directory

  • --ref-link give a reference link

  • --ref-link-prefix give a link prefix

  • --add_handle add a handle for linking to the page

  • --page_handle the page handle to use [default page name]

  • --add_heading add a heading to the page (uses the page name)

Clang-tidy

Clang-tidy is a set of tools which allows a developer to detect and fix code which does not follow current best practices, such as unused parameters or not using range-based for loops. Primarily this is used for modernising C++ code.

The full list of clang-tidy checks can be seen here.

Installing

Mantid does not come packaged with clang-tidy; each developer must download it themselves. Windows users can utilise clang-tidy support for Visual Studio. For other operating systems, Mantid provides clang-tidy functionality through cmake.

• Ubuntu: Run sudo apt-get install clang-tidy in the command line.

• Windows: Download the Visual Studio extension. Windows can operate clang-tidy from Visual Studio alone and so do not need to touch cmake.

For non-Ubuntu systems, download the latest clang-tidy pre-compiled binary. Windows users should add to path when prompted.

Visual Studio

Once you have installed the clang-tidy extension, Visual Studio will have additional options under Tools -> Options -> Clang Power Tools. Here you can select which checks to run via a checkbox or by supplying a custom check string. For example:

-*,modernize-*

will disable basic default checks (-*) and run all conversions to modern C++ (modernize-*), such as converting to range-based for loops or using the auto keyword. Other settings should not require alteration for clang-tidy to work.

To run clang-tidy, right click on a target and highlight Clang Power Tools. You will have a number of options. Tidy will highlight code which fails one of the checks, whereas Tidy fix will automatically change your code.

Note: clang-tidy does not work on ALL BUILD so it is necessary to select a subtarget.

Cmake

In the cmake gui, find the CLANG_TIDY_EXE parameter. If you are a non-Linux developer, you may have to manually point to your clang-tidy install. Configure, and check the cmake log for the message clang-tidy found. If the clang-tidy not found warning was posted instead then it has not worked.

Once you have clang-tidy, there are several relevant parameters you will want to change:

• ENABLE_CLANG_TIDY will turn on clang-tidy support.

• CLANG_TIDY_CHECKS is a semi-colon separated list of checks for clang-tidy to carry out. This defaults to all modernize- checks.

• APPLY_CLANG_TIDY_FIX will automatically change the code whenever a check has returned a result. The behaviour of ENABLE_CLANG_TIDY without this checked is to highlight issues only.

Configure the build to check that your selected options are reflected in CMAKE_CXX_CLANG_TIDY, and then generate. When you next build, clang-tidy will perform the selected checks on the code included in the target.

Note: There is a known issue that clang-tidy is only being applied to certain directories within Framework and Mantidplot.

Options

Some clang-tidy checks have optional arguments. For example, modernize-loop-convert, which changes loops to range-based, assigns a riskiness to each loop and can accept a MinConfidence argument to determine which risk levels to address.

Adding the optional arguments is clunky and will rarely be required, so it has not been directly added to Mantid’s cmake setup. To add optional arguments, add the following onto the end of CLANG_TIDY_CHECKS:

;-config={CheckOptions: [ {key: check-to-choose-option.option-name, value: option-value} ]}

For example, to convert all loops classified as risky or above, we would append:

;-config={CheckOptions: [ {key: modernize-loop-convert.MinConfidence, value: risky} ]}