Mercurial contains a simple regression test framework that allows both Python unit tests and shell-script driven regression tests.
See also: DebuggingTests
Running the test suite
To run the tests, do:
$ make tests cd tests && ./run-tests.py ............................................ Ran 44 tests, 0 failed.
This finds all scripts in the tests/ directory named test-* and executes them. The scripts can be either shell scripts or Python. Each test is run in a temporary directory that is removed when the test is complete.
You can also run tests individually:
$ cd tests/ $ ./run-tests.py test-pull test-undo .. Ran 2 tests, 0 failed.
A test-<x> succeeds if the script returns success and its output matches test-<x>.out. If the new output doesn't match, it is stored in test-<x>.err.
Also, run-tests.py has some useful options:
$ ./run-tests.py --help usage: run-tests.py [options] [tests] options: -h, --help show this help message and exit -v, --verbose output verbose messages -t TIMEOUT, --timeout=TIMEOUT kill errant tests after TIMEOUT seconds -c, --cover print a test coverage report -s, --cover_stdlib print a test coverage report inc. standard libraries -C, --annotate output files annotated with coverage -r, --retest retest failed tests -f, --first exit on the first test failure -R, --restart restart at last error -i, --interactive prompt to accept changed output
Running tests under Windows is a bit harder, see WindowsTestingPlan for details.
Note that tests won't run properly with an egg based install of Mercurial; the system install of Mercurial will be used instead of the checked out version. Use a Mercurial installed from source instead to avoid conflicts.
Writing a shell script test
Creating a regression test is easy. Simply create a shell script that executes the necessary commands to exercise Mercurial.
Here's an example:
hg init touch a hg add a hg commit -m "Added a" -d "0 0" touch main hg add main hg commit -m "Added main" -d "0 0" hg checkout 0 echo Main should be gone ls
Then run your test:
$ ./run-tests.py test-example . test-example generated unexpected output: Main should be gone a Ran 1 tests, 1 failed.
Double-check your script's output, then save the output so that future runs can check for the expected output:
$ mv test-example.err test-example.out $ ./run-tests.py test-example . Ran 1 tests, 0 failed.
Writing a Python unit test
A unit test operates much like a regression test, but is written in Python. Here's an example:
1 #!/usr/bin/env python 2 3 import sys 4 from mercurial import bdiff, mpatch 5 6 def test1(a, b): 7 d = bdiff.bdiff(a, b) 8 c = a 9 if d: 10 c = mpatch.patches(a, [d]) 11 if c != b: 12 print "***", `a`, `b` 13 print "bad:" 14 print `c`[:200] 15 print `d` 16 17 def test(a, b): 18 print "***", `a`, `b` 19 test1(a, b) 20 test1(b, a) 21 22 test("a\nc\n\n\n\n", "a\nb\n\n\n") 23 test("a\nb\nc\n", "a\nc\n") 24 test("", "") 25 test("a\nb\nc", "a\nb\nc") 26 test("a\nb\nc\nd\n", "a\nd\n") 27 test("a\nb\nc\nd\n", "a\nc\ne\n") 28 test("a\nb\nc\n", "a\nc\n") 29 test("a\n", "c\na\nb\n") 30 test("a\n", "") 31 test("a\n", "b\nc\n") 32 test("a\n", "c\na\n") 33 test("", "adjfkjdjksdhfksj") 34 test("", "ab") 35 test("", "abc") 36 test("a", "a") 37 test("ab", "ab") 38 test("abc", "abc") 39 test("a\n", "a\n") 40 test("a\nb", "a\nb") 41 42 print "done"
Writing Windows-only tests
Sometimes, it is necessary to write tests which will only run on Windows (for example, testing case sensitivity issues, or cases where os.sep is not '/'). The simplest way of doing this is to write the test as a .bat file. As usual, the output will be compared with the expected output, stored in a file with the .out extension.
Here is a simple example:
@echo off call hg init echo hello >a call hg add a call hg status
Some things to note:
Use @echo off at the start of the file to stop commands being echoed. (Although command echo could be useful to see what commands produce what output - just be careful to add the command echo into the expected output!)
You have to use call hg to run Mercurial. If you don't, the test will terminate after the first hg call!. This is because in the test environment, the hg command is implemented as a .bat file, and Windows won't allow one batch file to call another without an explicit call command. Without the call, control is transferred to the hg.bat file, and does not return.
Windows has a much more limited set of utilities available by default - so don't assume that things like sed exist. (If you're following WindowsTestingPlan, you will have many of the normal Unix tools, so in practice you can get away with this without too much bother).
Making Tests Repeatable
There are some tricky points here that you should be aware of when writing tests:
- hg commit wants user interaction - use -m "text"
- hg up -m wants user interaction, set HGMERGE to something noninteractive:
cat <<'EOF' > merge #!/bin/sh echo merging for `basename $1` EOF chmod +x merge env HGMERGE=./merge hg update -m 1
- changeset hashes will change based on user and date which make
- things like hg history output change - use -d:
hg commit -m "test" -u test -d "0 0"
- diff will show the current time - strip with sed:
hg diff | sed "s/\(\(---\|+++\) [a-zA-Z0-9_/.-]*\).*/\1/"
Making tests portable
You also need to be careful that the tests are portable from one platform to another. You're probably working on Linux, where the GNU toolchain has more (or different) functionality than on MacOS, *BSD, Solaris, AIX, etc. While testing on all platforms is the only sure-fire way to make sure that you've written portable code, here's a list of problems that have been found and fixed in the tests. Another, more comprehensive list may be found in the GNU Autoconf manual, online here:
The Bourne shell is a very basic shell. /bin/sh on Linux is typically bash, which even in Bourne-shell mode has many features that Bourne shells on other Unix systems don't have (and even on Linux /bin/sh isn't guaranteed to be bash). You'll need to be careful about constructs that seem ubiquitous, but are actually not available in the least common denominator. While using another shell (ksh, bash explicitly, posix shell, etc.) explicitly may seem like another option, these may not exist in a portable location, and so are generally probably not a good idea. You may find that rewriting the test in python will be easier.
- don't use pushd/popd; save the output of "pwd" and use "cd" in place of the pushd, and cd back to the saved pwd instead of popd.
- don't use math expressions like let, (( ... )), or $(( ... )); use "expr" instead.
- don't use the -q option; redirect stdout to /dev/null instead.
- don't use extended regular expressions with grep; use egrep instead, and don't escape any regex operators.
- make sure that the beginning-of-line matcher ("^") is at the very beginning of the expression -- it may not be supported inside parens.
- don't use the -i option; instead, redirect to a file:
sed -e 's/foo/bar/' a > a.new mv a.new a
- echo may interpret "\n" and print a newline; use printf instead if you want a literal "\n" (backslash + n).
- false is guaranteed only to return a non-zero value; you cannot depend on it being 1. On Solaris in particular, /bin/false returns 255. Rewrite your test to not depend on a particular return value, or create a temporary "false" executable, and call that instead.
- don't use the -N option. There's no particularly good workaround short of writing a reasonably complicated replacement script, but substituting gdiff for diff if you can't rewrite the test not to need -N will probably do.
- don't use it, or else eliminate leading whitespace from the output:
wc -l | sed -e 's/^ *//'
or use python:
python -c "print len(open('foo').readlines())"
- don't use the -c option (not part of SUSv3, not supported on OpenBSD). Instead, use dd. the following are equivalent; the latter is preferred:
head -c 20 foo > bar
dd if=foo of=bar bs=1 count=20 2>/dev/null