On Tue, Nov 29, 2016 at 7:25 AM, Petr Gajdos <pgaj...@suse.cz> wrote:
> On Tue, Nov 29, 2016 at 01:09:25PM +0000, Joe Orton wrote: > > On Mon, Nov 28, 2016 at 05:16:12PM -0600, William A Rowe Jr wrote: > > > httpd: Syntax error on line 295 of > > > /home/wrowe/dev/test/test24-apr16-ossl102/t/conf/httpd.conf: Cannot > load > > > /home/wrowe/dev/test/test24-apr16-ossl102/c-modules/test_ > session/.libs/mod_test_session.so > > > into server: > > > /home/wrowe/dev/test/test24-apr16-ossl102/c-modules/test_ > session/.libs/mod_test_session.so: > > > undefined symbol: ap_hook_session_save > > > > > > Does anyone have an uncommitted test framework patch > > > up their sleeve to remedy? > > > > Modules using the session API can't be loaded if mod_session isn't > > loaded. > Of course. Note this threw me as mod_test_ssl doesn't suffer this same issue, since it correctly relied on optional hooks to avoid load order dependency (and emits errors in the log that it can't function if mod_ssl isn't loaded.) > > To make this "work" in the test framework we'd need to surround the > > generated LoadModule for mod_test_session with <IfModule mod_session.c> > > or something. I don't know if that can be done without hacking around > > inside Apache::Test. > > I was thinking about text file along each c-module which would express > the required modules which would be translated to <IfModule></IfModule>. > It seems like something inside Apache::Test that indicated module dependencies for compiled test modules would be worthwhile. Looked some this morning, and noticed no special sauce other than #define HTTPD_TEST_REQUIRE_APACHE 2 which isn't quite the same thing. But it does look like any special sauce should be added to Apache-Test/lib/Apache/TestConfigC.pm unless we do some <IfModule> dependency logic. (Or simply fixed mod_session to not be load order dependent and use the optional entry point APIs.)
