FYI,
I dug a little deeper wrt to the test failures in the Ruby bindings.
Comparing the code generated with swig 1.31 and 1.33,
e.g. ./subversion/bindings/swig/ruby/core.c,
swig 1.33 added lines of the form
vresult = rb_ary_new();
to lots of wrapper functions. If I comment out these and rebuild,
the Ruby test cases succeed.
Looking at the swig SVN repository, these additional lines
are the result of commit 9746. Its log says:
Fixed a bug with argout values being incorrectly
returned when the first argout/returned element
was an array.
That doesn't look correct to me, e.g one wrapper that
has the above line is for svn_opt_format_option
which doesn't have "out" parameters that are arrays,
it's only "out" parameter is a string:
/**
* Print an option @a opt nicely into a @a string allocated in @a pool.
* If @a doc is set, include the generic documentation string of @a opt,
* localized to the current locale if a translation is available.
*/
void
svn_opt_format_option(const char **string,
const apr_getopt_option_t *opt,
svn_boolean_t doc,
apr_pool_t *pool);
Cheers, Roderich
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
