On 20.09.2017 00:29, brane_at_apache.org wrote:
> Author: brane
> Date: Tue Sep 19 22:29:39 2017
> New Revision: 1808955
>
> URL: http://svn.apache.org/viewvc?rev=1808955&view=rev
> Log:
> The compile-time and run-time version of httpd may not be the same.
> Since the tests rely on it to interpret test case results, make sure
> that they see the run-time version.
>
> * subversion/tests/cmdline/davautocheck.sh
> (HTTPD_VERSION): New; get it from apxs.
> Override the httpd version for 'make check' etc.
>
> Modified:
> subversion/trunk/subversion/tests/cmdline/davautocheck.sh
>
> Modified: subversion/trunk/subversion/tests/cmdline/davautocheck.sh
> URL: http://svn.apache.org/viewvc/subversion/trunk/subversion/tests/cmdline/davautocheck.sh?rev=1808955&r1=1808954&r2=1808955&view=diff
> ==============================================================================
> --- subversion/trunk/subversion/tests/cmdline/davautocheck.sh (original)
> +++ subversion/trunk/subversion/tests/cmdline/davautocheck.sh Tue Sep 19 22:29:39 2017
> @@ -778,14 +778,16 @@ else
> || fail "Subversion client couldn't find and/or load ra_dav library '$HTTP_LIBRARY'"
> fi
>
> +HTTPD_VERSION="$($APXS -q HTTPD_VERSION)"
> +
> if [ $# = 0 ]; then
> - TIME_CMD "$MAKE" check "BASE_URL=$BASE_URL" $SSL_MAKE_VAR
> + TIME_CMD "$MAKE" check "BASE_URL=$BASE_URL" "HTTPD_VERSION=$HTTPD_VERSION" $SSL_MAKE_VAR
> r=$?
> else
> (cd "$ABS_BUILDDIR/subversion/tests/cmdline/"
> TEST="$1"
> shift
> - TIME_CMD "$ABS_SRCDIR/subversion/tests/cmdline/${TEST}_tests.py" "--url=$BASE_URL" $SSL_TEST_ARG "$@")
> + TIME_CMD "$ABS_SRCDIR/subversion/tests/cmdline/${TEST}_tests.py" "--url=$BASE_URL" "--httpd-version=$HTTPD_VERSION" $SSL_TEST_ARG "$@")
> r=$?
> fi
I think this should fix the test failures on OSX 10.9. The problem
there, as I mentioned in the other thread, is that configure gets httpd
version 2.2.26 from the header files, but the actual version reported by
the server is 2.2.29. The former has an URL quoting bug that we test for
in basic_tests.py, the latter doesn't.
I believe our configure and test scripts were written with the
assumption that the compile-time and run-time versions of httpd are the
same. That turns out not to (always) be the case, even in the most
trivial cases.
If anyone objects to this change, speak now or forever hold your piece. :)
-- Brane
Received on 2017-09-20 00:34:36 CEST