Canonify outputs when updating baselines via btest-diff

This leverages the same canonicalization btest-diff already applies
during test output comparison against baselines also when updating
those baselines (usually via btest -U/-u). This removes a bunch of
noise from the baselines, including personal home directories,
timestamps, etc.

Since btest-diff doesn't know whether the baseline has undergone this
canonicalization, it continues to canonicalize the basline prior to
diffing. To track whether it has canonicalized a baseline when
updating, btest-diff now also prepends a header to the generated
baseline that warns users about the fact that it is
auto-generated. The presence of this header doubles as a marker for
canonicalization.

Includes a test-case.

btest-diff

@@ -16,6 +16,41 @@
 # Maximum number of lines to show from mismatching input file by default.
 MAX_LINES=100
 
+# Header line we tuck onto new baselines generated by
+# btest-diff. Serves both as a warning and as an indicator that the
+# baseline has been run through the TEST_DIFF_CANONIFIER (if any).
+HEADER="### btest baseline data generated by btest-diff. Do not edit. Use btest -U/-u to update."
+
+# Predicate, succeeds if the given baseline is canonicalized.
+function is_canon_baseline {
+    local input="$1"
+
+    if head -1 "$input" | grep -q -F "$HEADER" 2>/dev/null; then
+        return 0
+    fi
+
+    return 1
+}
+
+# Prints the requested baseline to standard out if it is
+# canonicalized. Otherwise, fails.
+function get_canon_baseline {
+    local input="$1"
+
+    ! is_canon_baseline "$input" && return 1
+    tail -n +2 "$input"
+}
+
+# Updates the given baseline to the given destination, prepending our
+# header.
+function update_canon_baseline {
+    local input="$1"
+    local output="$2"
+
+    echo "$HEADER" >"$output"
+    cat "$input" >>"$output"
+}
+
 if [ -n "$TEST_DIFF_FILE_MAX_LINES" ]; then
     MAX_LINES=$TEST_DIFF_FILE_MAX_LINES
 fi
@@ -67,34 +102,43 @@ fi
 if [ -e $TEST_BASELINE/$canon ]; then
     error=0
 
-    if [ "$TEST_DIFF_CANONIFIER" != "" ]; then
-        diff1=/tmp/test-diff.$$.$canon.baseline.tmp
-        diff2=/tmp/test-diff.$$.$canon.tmp
-        tmpfiles="$tmpfiles $diff1 $diff2"
+    # If no canonifier is defined, make it a runnable
+    # no-op. Simplifies code layout.
+    if [ -z "$TEST_DIFF_CANONIFIER" ]; then
+        TEST_DIFF_CANONIFIER=cat
+    fi
 
-        eval $TEST_DIFF_CANONIFIER $TEST_BASELINE/$canon <$TEST_BASELINE/$canon >$diff1
+    canon_baseline=/tmp/test-diff.$$.$canon.baseline.tmp
+    canon_output=/tmp/test-diff.$$.$canon.tmp
+    tmpfiles="$tmpfiles $canon_baseline $canon_output"
+
+    # Prepare the baseline. When created by a recent btest-diff, we
+    # don't need to re-canonicalize, otherwise we do.
+    if ! get_canon_baseline $TEST_BASELINE/$canon >$canon_baseline; then
+        # It's an older uncanonicalized baseline, so canonicalize
+        # it now prior to comparison. Future updates via btest
+        # -U/-u will then store it canonicalized.
+        eval $TEST_DIFF_CANONIFIER <$TEST_BASELINE/$canon >$canon_baseline
         if [ $? -ne 0 ]; then
             error=1
             echo "== Error ==============================" >>$TEST_DIAGNOSTICS
             echo "btest-diff: TEST_DIFF_CANONIFIER failed on file '$TEST_BASELINE/$canon'" >>$TEST_DIAGNOSTICS
             result=1
         fi
+    fi
 
-        eval $TEST_DIFF_CANONIFIER $input <$input >$diff2
-        if [ $? -ne 0 ]; then
-            error=1
-            echo "== Error ==============================" >>$TEST_DIAGNOSTICS
-            echo "btest-diff: TEST_DIFF_CANONIFIER failed on file '$input'" >>$TEST_DIAGNOSTICS
-            result=1
-        fi
-    else
-        diff1=$TEST_BASELINE/$canon
-        diff2=$input
+    # Canonicalize the new test output
+    eval $TEST_DIFF_CANONIFIER $input <$input >$canon_output
+    if [ $? -ne 0 ]; then
+        error=1
+        echo "== Error ==============================" >>$TEST_DIAGNOSTICS
+        echo "btest-diff: TEST_DIFF_CANONIFIER failed on file '$input'" >>$TEST_DIAGNOSTICS
+        result=1
     fi
 
     if [ $error -eq 0 ]; then
         echo "== Diff ===============================" >>$TEST_DIAGNOSTICS
-        diff -au $@ $diff1 $diff2 >>$TEST_DIAGNOSTICS
+        diff -au $@ $canon_baseline $canon_output >>$TEST_DIAGNOSTICS
         result=$?
     fi
 else
@@ -119,15 +163,20 @@ elif [ "$TEST_MODE" == "UPDATE_INTERACTIVE" ]; then
 
     echo -n "$TEST_NAME ..." >/dev/tty
 
+    # If we have a canonified version of the input file, use it as the
+    # new baseline. Otherwise, just use the input.
+    update="${canon_output:-$input}"
+
     if [ $rc == 0 ]; then
-        cp $input $TEST_BASELINE/$canon
+        update_canon_baseline $update $TEST_BASELINE/$canon
         exit 0
     fi
 
     exit $rc
 
 elif [ "$TEST_MODE" == "UPDATE" ]; then
-    cp $input $TEST_BASELINE/$canon
+    update="${canon_output:-$input}"
+    update_canon_baseline $update $TEST_BASELINE/$canon
     exit 0
 fi
 

testing/tests/canonifier-conversion.test

@@ -0,0 +1,47 @@
+# This test verifies that we only canonicalize baselines once, namely
+# when canonicalizing new test output, and that baselines get
+# converted over to our header-tagged format.
+#
+# Test prep: make our canonifier executable
+# %TEST-EXEC: chmod 755 ./diff-double-x
+#
+# Test prep: create a "dated" baseline.
+# %TEST-EXEC: mkdir -p Baseline/canonifier-conversion
+# %TEST-EXEC: echo x >Baseline/canonifier-conversion/output
+#
+# Verify that it succeeds:
+# %TEST-EXEC: btest %INPUT
+#
+# Update the baseline. This should prefix btest-diff's header and canonify:
+# %TEST-EXEC: btest -U %INPUT
+#
+# Verify that these have happened and preserve baseline:
+# %TEST-EXEC: head -1 Baseline/canonifier-conversion/output | grep -q 'Do not edit'
+# %TEST-EXEC: tail -1 Baseline/canonifier-conversion/output | grep xx
+# %TEST-EXEC: cp Baseline/canonifier-conversion/output base.1
+#
+# Verify that it still succeeds:
+# %TEST-EXEC: btest %INPUT
+#
+# Update the baseline again.
+# %TEST-EXEC: btest -U %INPUT
+# %TEST-EXEC: cp Baseline/canonifier-conversion/output base.2
+#
+# Verify the updated baselines remained identical.
+# %TEST-EXEC: test "$(cat base.1)" = "$(cat base.2)"
+
+@TEST-EXEC: echo x >output
+@TEST-EXEC: btest-diff output
+
+%TEST-START-FILE btest.cfg
+[btest]
+TmpDir      = .tmp
+
+[environment]
+TEST_DIFF_CANONIFIER=%(testbase)s/diff-double-x
+%TEST-END-FILE
+
+%TEST-START-FILE diff-double-x
+#! /usr/bin/env bash
+sed 's/x/xx/g'
+%TEST-END-FILE