test: make test_expect_equal_file() arguments flexible
authorDmitry Kurochkin <dmitry.kurochkin@gmail.com>
Wed, 1 Feb 2012 07:19:54 +0000 (11:19 +0400)
committerDavid Bremner <bremner@debian.org>
Sun, 2 Sep 2012 02:09:26 +0000 (23:09 -0300)
Before the change, test_expect_equal_file() function treated the first
argument as "actual output file" and the second argument as "expected
output file".  When the test fails, the files are copied for later
inspection.  The first files was copied to "$testname.output" and the
second file to "$testname.expected".  The argument order for
test_expect_equal_file() is often wrong which results in confusing
diff output and incorrectly named files.

The patch solves the issue by changing test_expect_equal_file() to
treat arguments just as two files, without any special properties
(like "actual" and "expected").  The file names for copying is now
based on the given file name: "$testname.$file1" and
"$testname.$file2".  E.g. if test_expect_equal_file() is called with
"OUTPUT" and "EXPECTED", the copied files can be named
"emacs.1.OUTPUT" and "emacs.1.EXPECTED".

The down side of this approach is that diff argument order depends on
test_expect_equal_file() argument order.  So sometimes we get diff
from expected to actual results, and sometimes the other way around.
But the files are always named correctly.

test/README
test/test-lib.sh

index 43656a35baf9f9bc5e5d41aca2b65c8e4e763450..e0364e86652e6422c8172894d4dc16436ec06adc 100644 (file)
@@ -176,12 +176,12 @@ library for your script to use.
    will generate a failure and print the difference of the two
    strings.
 
- test_expect_equal_file <output> <expected>
+ test_expect_equal_file <file1> <file2>
 
-   Identical to test_exepect_equal, except that <output> and
-   <expected> are files instead of strings.  This is a much more
-   robust method to compare formatted textual information, since it
-   also notices whitespace and closing newline differences.
+   Identical to test_exepect_equal, except that <file1> and <file2>
+   are files instead of strings.  This is a much more robust method to
+   compare formatted textual information, since it also notices
+   whitespace and closing newline differences.
 
  test_debug <script>
 
index 791d2dc668e05dde0a0c4e5828db1dea8988e2b6..663b18e69d05558fc624aef1cd4c5166977c240a 100644 (file)
@@ -497,17 +497,17 @@ test_expect_equal_file ()
        test "$#" = 2 ||
        error "bug in the test script: not 2 or 3 parameters to test_expect_equal"
 
-       output="$1"
-       expected="$2"
+       file1="$1"
+       file2="$2"
        if ! test_skip "$test_subtest_name"
        then
-               if diff -q "$expected" "$output" >/dev/null ; then
+               if diff -q "$file1" "$file2" >/dev/null ; then
                        test_ok_ "$test_subtest_name"
                else
                        testname=$this_test.$test_count
-                       cp "$output" $testname.output
-                       cp "$expected" $testname.expected
-                       test_failure_ "$test_subtest_name" "$(diff -u $testname.expected $testname.output)"
+                       cp "$file1" "$testname.$file1"
+                       cp "$file2" "$testname.$file2"
+                       test_failure_ "$test_subtest_name" "$(diff -u "$testname.$file1" "$testname.$file2")"
                fi
     fi
 }