![John Ericson](/assets/img/avatar_default.png)
When we pipe to `>(...)` like that, we unfortunately don't wait for the process to finish. Better to just substitute the file. Also, use the "unified" diff output that people (including myself) are more familiar with, thanks to Git.
34 lines
1,010 B
Bash
34 lines
1,010 B
Bash
# Golden test support
|
|
#
|
|
# Test that the output of the given test matches what is expected. If
|
|
# `_NIX_TEST_ACCEPT` is non-empty also update the expected output so
|
|
# that next time the test succeeds.
|
|
function diffAndAcceptInner() {
|
|
local -r testName=$1
|
|
local -r got="$2"
|
|
local -r expected="$3"
|
|
|
|
# Absence of expected file indicates empty output expected.
|
|
if test -e "$expected"; then
|
|
local -r expectedOrEmpty="$expected"
|
|
else
|
|
local -r expectedOrEmpty=lang/empty.exp
|
|
fi
|
|
|
|
# Diff so we get a nice message
|
|
if ! diff --unified "$got" "$expectedOrEmpty"; then
|
|
echo "FAIL: evaluation result of $testName not as expected"
|
|
badDiff=1
|
|
fi
|
|
|
|
# Update expected if `_NIX_TEST_ACCEPT` is non-empty.
|
|
if test -n "${_NIX_TEST_ACCEPT-}"; then
|
|
cp "$got" "$expected"
|
|
# Delete empty expected files to avoid bloating the repo with
|
|
# empty files.
|
|
if ! test -s "$expected"; then
|
|
rm "$expected"
|
|
fi
|
|
fi
|
|
}
|