On Thursday 02 February 2012, Joe Orton wrote: > The combination of APR_SUCCESS and DECLINED is unusual; an int > return value with OK/DECLINED?
Input and output filters should return an apr_status_t. So, if the hook does not return an apr_status_t, core_input_filter() would have to invent some apr_status_t value, which is bad. And I think the general principle that return code 'int' means HTTP_* error code should be kept. But I agree that DECLINED == -1 is not a good idea, because it could in theory collide with another APR_E* code. Should we define an AP_STATUS_DECLINED or something in the APR_OS_START_USERERR range? Or simply change the definition of DECLINED to be in the APR_OS_START_USERERR range? APR_OS_START_USERERR is 120000, so there should be no danger of a collision with HTTP_*. It would require a major MMN bump, but making core_output_filter_ctx_t and core_ctx_t private is an API change, anyway.
