I tested the converter from the legacy (0.9.23) format to the new format and I think I've found a bug, but I'm unsure how to best fix it. Maybe Colin (who is more into the format) would know better? The problem: entry_read (in module-device-restore.c) is called from firing the PA_CORE_HOOK_SINK_NEW hook (see sink_new_hook_callback). Entry_read reads the old sink entry successfully and tries to iterate sink->ports. However, sink->ports is garbage is that point (as the sink is allocated with xmalloc, not xmalloc0), and so segfault is possible. sink->ports is not set until after the PA_CORE_HOOK_SINK_NEW hook has finished. Possible solutions include e g setting sink->ports earlier as well as listening to pa_sink_fixate instead of pa_sink_new, but I wonder if any of them may have unwanted side effects? Or maybe save a temporary null-port entry that is used later (as a fallback is there is no port entry)? -- David Henningsson, Canonical Ltd. http://launchpad.net/~diwic
