Markus Schaber <[email protected]> writes: > Hi, Philip, > > I did address most of the issues. However, the following question is open: > > Von: Philip Martin [[email protected]] >> Markus Schaber <[email protected]> writes: > >>> +/* Function to prepare a single test file */ >>> + >>> +static svn_error_t * >>> +create_test_file(struct test_file_definition_t* definition, apr_pool_t >>> *pool, apr_pool_t *scratch_pool) >>> +{ >>> + apr_status_t status; >>> + apr_file_t *file_h; >>> + apr_off_t midpos = definition->size / 2; >>> + svn_error_t *err = NULL; >>> + int i; >>> + >>> + if (definition->size < 5) >>> + SVN_ERR_ASSERT(strlen(definition->data) >= definition->size); >>> + else >>> + SVN_ERR_ASSERT(strlen(definition->data) >= 5); >>> + >>> + status = apr_filepath_merge(&definition->created_path, >>> + TEST_DIR, >>> + definition->name, >>> + 0, >>> + pool); > >> Do you need to do that? Could you simply relative paths as is done >> subversion/tests/libsvn_diff/diff-diff3-test.c. > > I used path concatenation to pack all of the files into one temporary > directory, so they can be easily cleaned up, and only one svn:ignore > entry is needed. I did model this after how the temporary > repositories are handled in svn_test_fs.c.
I meant use svn_relpath_join rather than apr_filepath_merge. We generally prefer the svn_ functions over the apr_ functions. The svn_ functions often lead to simpler error handling because they return svn_error_t. The svn_ IO functions also take paths in UTF-8 which usually makes things easier since paths are generally UTF-8 in Subversion code. Putting the files in one directory is good because the directory can be added to the cleanup list. -- uberSVN: Apache Subversion Made Easy http://www.uberSVN.com
