Skip to content

Latest commit

 

History

History
105 lines (83 loc) · 4.84 KB

README.md

File metadata and controls

105 lines (83 loc) · 4.84 KB
Raw

vim-markbar Tests

Tests are written using vader.vim and are run with the run_tests.sh script. This script can only be run from this test directory.

See ./run_tests --help for options.

Some common test workflows are given below:

# from project root,
cd test
./run_tests.sh           # to run test cases in vim
./run_tests.sh --neovim  # to run test cases in neovim

# run all tests in a non-headless, visible vim instance
./run_tests.sh -v

# ditto, but in neovim
./run_tests.sh -v --neovim

# run all tests in English, and again in an international locale
./run_tests.sh --international

# run a specific test in a visible vim instance
./run_tests.sh -v -f test-MarkbarModel.vader

# run a set of sequential standalone tests
./run_tests.sh -v -f standalone-test-sequential-name-persistence.vader

# To run test cases in a visible, non-headless neovim instance,
# using /usr/local/bin/nvim as the nvim executable,
# *Note* that run_tests.sh doesn't know if a vim/nvim executable is passed
# as the --vim_exe, so you still need to omit/specify --neovim.
./run_tests.sh --neovim -v --vim_exe=/usr/local/bin/nvim

Test Organization

By default, vader.vim doesn't isolate tests from each other: global variables set in one test will persist when other tests are run. See vader.vim #101.

For this reason, we maintain three kinds of tests: "regular" tests, standalone tests, and standalone sequential tests.

"Regular" tests are named like test-specificthing.vader or test-ClassName.vader. When running all tests (by omitting -f arguments to run_tests.sh), "regular" tests are globbed together and run in a single (n)vim instance.

Standalone tests are named like standalone-test-specificthing.vader. run_tests.sh always runs each of these tests in a new (n)vim instance. Exiting and restarting (n)vim between tests ensures that these tests are isolated from each other. Roughly speaking, standalone tests are tests that manipulate vim or vim-markbar settings.

Standalone sequential tests are tests that require a viminfo/shada file. Regular and standalone tests pass -i NONE to (n)vim, to prevent viminfo/shada from affecting test execution. Some of vim-markbar's features rely on viminfo/shada, like persistent user-given mark names, and these can't be tested without a viminfo/shada file.

Standalone sequential tests are kept in subdirectories named like standalone-test-sequential-something.vader, and are named like XX-standalone-test-sequential-something.vader within that subdirectory, where XX is a number (with leading 0, if applicable). Standalone sequential tests in a subdirectory are each run in separate (n)vim instances, one after the other in ascending numerical order, and share the same initially empty viminfo/shada file with each other.

A standalone sequential test need not exclusively consist of *.vader files. One can have a standalone sequential test with the folder structure:

standalone-test-sequential-something.vader/01-standalone-test-sequential-something.vader
standalone-test-sequential-something.vader/02-standalone-test-sequential-something.sh
standalone-test-sequential-something.vader/03-standalone-test-sequential-something.vader

In this folder, the 01-standalone-test-sequential-something.vader test suite will run first, and then the 02-standalone-test-sequential-something.sh bash script will run, followed by the 03-standalone-test-sequential-something.vader test suite.

Bash scripts in a standalone sequential test must take, as their first argument, the filepath of the vim executable being used to run the test; and as their second argument, "vim arguments" such as -Nnu .test_vimrc. See standalone-test-sequential-filetype.vader for an example. If the bash script returns a nonzero exit code, the test will fail.

Other Notes and Known Issues

vader.vim is sensitive to trailing whitespace in its Expect: blocks, where they denote an empty line in the buffer. Some tests (particularly standalone-test-openmarkbar.vader and standalone-test-peekaboo.vader) rely on this behavior. If you use vim autocommands to automatically delete trailing whitespace on buffer write, then you should disable those autocommands when editing vader files.

Test output from running run_tests.sh with a vim executable is significantly slower than with neovim and is badly mangled because the running vim instance doesn't respect the actual size of the terminal window. See Issue #53.