AutoTest setup
AutoTest is fully flexible to work with any programming language, workflow, tool and library you want to use.
To make CodeGrade AutoTest as flexible as possible, it is possible to customize the complete environment the tests are run in. Each separate assignment that makes use of AutoTest runs on its own Virtual Server in the cloud, to which you have superuser rights and network access during the setup phase. Every server on runs with the latest LTS version of Ubuntu, which is Ubuntu 18.04.2 LTS.
After the Setup Phase is finished, a snapshot is created which is used to initialize the containers used to run the actual tests on the student submissions. The Setup Phase consists of the default installed software, fixtures and setup script.
The location of the fixtures directory is random for every category and setup script to make it more virtually impossible for students to guess the location and changes. The location is stored and accessible via the$FIXTURESenvironment variable. Setup scripts and other scrips (fixtures) that are executed have to use Unix Line Endings (LF) and not Windows Line Endings (CRLF)! There are multiple options for setting up your environment, which makes AutoTest easy to use for simple cases yet very flexible for all advanced cases. ## Default Installed Software Each AutoTest VPS environment comes with pre-installed software that is commonly used for testing and running student submissions. This body of software is sufficient for most cases, which allows you to skip manually further setting up the environment. The following software is automatically installed in all environments, all versions, except for Python, are lower bounds, as all packages are always updated to the latest version shipped by Ubuntu: • Python 2.7 (with pip) • Python 3.6 (with pip3) • Python 3.7 (with pip3) • Java 8 (openjdk-8-jdk) • Java 11 (openjdk-11-jdk) • Jupyter/IPython • Mono (6.4) • .NET 6.0 • ASP.NET 6.0 • Node (JavaScript 8.10) • Octave (4.2.2) • R (4.1.3) • C/C++ (gcc 7 and clang 6 as compilers) • Go (golang 1.10) • Git • Maven • Flake8 • Numpy (for Python2, Python3.6 and Python3.7) • SciPy (for Python2, Python3.6 and Python3.7) • Check (unit test framework for C) Read the following sections to find out about extending this environment with other required software. ### Unit testing wrapper scripts In addition to the software listed above, CodeGrade provides custom wrapper scripts for popular unit testing frameworks to make integration with Autotest a breeze. See the Unit Test chapter for a list of supported unit testing frameworks. ## Result visibility When a student uploads, AutoTest immediately runs on the handed in submission. You can choose whether you want the student to directly be able to view the AutoTest result (continuous feedback) or that the student can only view the result when the assignment is set to "Done" (just like with any other feedback). The AutoTest rubric categories will be filled in automatically and directly in either case if you have not created any hidden steps. If you have created hidden steps, AutoTest will be run again automatically within 30 minutes after the deadline, where the hidden steps are also executed. Before the deadline, hidden steps are never executed. ## Rubric calculation The rubric calculation setting determines how discrete rubric categories will be filled-in. ### Discrete Categories There are two modes to fill in discrete rubric categories: 1. 1. Minimum: A category's item will be chosen when the lower bound of this item is reached (e.g. when a category has 4 items and 75% of the tests succeed, the maximum item is filled in). 2. 2. Maximum: A category's item will be chosen when the upper bound of this item is reached (e.g. you need 100% passed tests to have the maximum item filled in). The percentages of the items in a category are independent of the amount of points given to them. E.g. if you have 4 items, item 1 is always 0%-25%, item 2 is 25%-50% and so forth. ### Continuous Categories Continuous rubric categories are filled in one way, this means this setting will have no influence on rubric calculation for this type of rubric category. The percentage achieved in the AutoTest level will also be the percentage achieved in the continuous rubric category. Example: You have a continuous rubric category with a maximum of 5 points, and an AutoTest level with a maximum of 10 points. If a user achieves 7 AutoTest points for this level, in the continuous rubric category the student will receive $5 \times \frac{7}{10} = 3.5$ points. ## Preferred Revision When the preferred revision is set to "Teacher", and a teacher's revision is available for a submission, AutoTest is run against the teacher revision instead of the code submitted by the student. If no teacher's revision is present AutoTest will be run against the code of the student. Setting the preferred revision to teacher can be useful if a student has made a tiny mistake in their code -- for example a misplaced punctuation mark -- that causes the majority of the tests to fail. The teacher can correct this mistake in the FileSystem and run the tests again to see what the score of this student would have been if such a mistake weren't made. After the teacher has made their changes, the AutoTest should be manually restarted if it has already started or finished, to make it run against the teacher's revision. You can restart an AutoTest by going to a result, clicking on the arrow next to the state of the result, and selecting "Restart this result". ## AutoTest caching AutoTest caching will cache the state of your AutoTest environment after your configuration (e.g. after downloading data, installing additional software or packages or unarchiving). When it is restarted, it can use this cached state to restart instantly and allow for very fast feedback to students. It is also very useful when iteratively developing your AutoTest, as changes in test steps will not require a cache refresh. AutoTest caching is turned on by default and should be left turned on for most configurations. There are however some uncommon reasons to disable it. Learn more in our guide here: ## Uploading fixtures Fixtures can be optionally uploaded to the AutoTest VPS. Fixtures are files you can upload prior to the test, which will be available in every separate test container. Use cases are files used as setup script, unit tests, custom software to run or install and test input. Select the fixtures to be uploaded and submit these to upload. A list of previously uploaded fixtures can be found above the upload dialog and managed here too. Uploading fixtures with the same name as current fixtures will by default result in an additional copy with (1), the 'Override duplicates' option can be used to overwrite current fixtures with the same name. Archives are not automatically extracted when uploading fixtures. This makes it possible to use unextracted archives as fixtures too. Use the commandstar xfvz$FIXTURES/ARCHIVE.tar.gzorunzip $FIXTURES/ARCHIVE.zipto extract archives manually. Be careful with the permissions, we recommend runningchown -R codegrade:codegrade$FIXTURES/dirandchmod -R 750 $FIXTURES/dirafter extracting. ### Limiting student access It is sometimes desirable to limit student access to fixtures or to limit the visibility of your uploaded fixtures. For instance if one of your fixtures is a solution to the assignment you use to test student submissions against. By default fixtures and their contents are visible to students. Toggle the eye icon next to the fixture to toggle its visibility to students inside the CodeGrade interface. Next to that, we offer multiple means of further limiting undesirable student access to fixtures. Firstly, the path to the fixtures is randomly generated for each category and thus only accessible using the $FIXTURES environment variable. This makes it harder for students to access the path, but not impossible.
A way to further limit student permissions in the $FIXTURES folder is to execute student code with the become_nobody command. When executed in this mode, students will have no permissions to read from the $FIXTURES folder. They will have permissions in the $STUDENT folder, which is the current directory in which student submission files are accessible, to read and execute. Copying files from the $FIXTURES directory to the $STUDENT directory with the cp or mv commands will not change permissions on these files, and the nobody user will not be able to read them. Use chmod 755 <FILE> to properly set these or use the install command to set these right away: install -m 755$FIXTURES/<fixture> $STUDENT. By default, scripts ran with the become_nobody command cannot write new files to the $STUDENT directory. Setting the write permission on the entire $STUDENT directory may be undesirable, as students may be able to overwrite their own code during the tests. Therefore, we recommend you create a new subdirectory where the output should be written with install -Dm 777$STUDENT/<SUBDIR>. If this subdirectory contains files that should not be read by students, use permission 733.

## Global setup script

A setup script can be specified which runs prior to the tests to customize the initial environment. Any script can be uploaded as fixture and subsequently run with the command given in the Global setup script to run input field.
This can be, for example, a bash script that installs software using apt and extracts archives, or compiles unit tests. Of course, one or multiple commands can also be given in the Global setup script to run directly.
If you need to setup or compile software for each student specifically and not globally, use the Per student setup script for this. Install any packages using the Global setup script as this will greatly increase the speed of AutoTest Runs
Network access and Superuser rights are available during the Setup Phase.
Setup scripts and other scrips (fixtures) that are executed have to use Unix Line Endings (LF) and not Windows Line Endings (CRLF)!

## Per student setup script

Use the per student setup script to compile, for example, each submission's code. This part is executed when setting up a student's individual environment and the current directory already holds the student's submission.
Compiling student code using the per student setup script has the benefit of the compiled code being available throughout all categories. If you want compiling to be part of a test, use the Run program test for this.