On Fri, 26 Jul 2019 15:43:07 +0300
"Tzvetomir Stoyanov (VMware)" <tz.stoyanov@xxxxxxxxx> wrote:
> --- a/lib/trace-cmd/trace-util.c
> +++ b/lib/trace-cmd/trace-util.c
> @@ -29,18 +29,6 @@
> int tracecmd_disable_sys_plugins;
> int tracecmd_disable_plugins;
>
> -static struct registered_plugin_options {
> - struct registered_plugin_options *next;
> - struct tep_plugin_option *options;
> -} *registered_options;
> -
> -static struct trace_plugin_options {
> - struct trace_plugin_options *next;
> - char *plugin;
> - char *option;
> - char *value;
> -} *trace_plugin_options;
Hmm, so this is identical to what's in lib/traceevent/event-plugin.c.
This doesn't affect this patch (and I may just take this patch as is),
but I'm thinking that we should make the plugins part of the tep
handler. That way, if you have two different tep handlers, each one
will need to register its own set of plugins.
We'll have to look to see if that's possible, because we will need to
do that for the application as a whole. Or perhaps we can add a
"tep_clone_plugin" option that allows a single "plugin" to be
registered to multiple tep handlers.
Again, this is a conversation not to really do with this patch, but
something to think about before releasing the libtraceevent library to
the wild.
-- Steve
|