On 20.08.21 г. 17:35, Steven Rostedt wrote:
On Fri, 20 Aug 2021 16:36:12 +0300 Yordan Karadzhov <y.karadz@xxxxxxxxx> wrote: if (field_mask & ignore_mask)+ continue; + trace_seq_printf(s, " %s=", field->name); _tep_print_field(s, data, field, &parse); - field = field->next; } }+void tep_print_selected_fields(struct trace_seq *s, void *data,As the above is an API, it needs a kernel doc type comment, and also an addition to the man pages. The man page may be a separate patch.I was thinking of maybe changing the second argument of the function to void tep_print_selected_fields(struct trace_seq *s, struct tep_record *record, struct tep_event *event, unsigned long long ignore_mask)-- Steve+ struct tep_event *event, + unsigned long long ignore_mask) +{ + print_selected_fields(s, data, event, ignore_mask);respectively here we will have print_selected_fields(s, record->data, event, ignore_mask);+}This way the call will look cleaner. tep_print_selected_fields(s, record, event, mask); instead of tep_print_selected_fields(s, record->data, event, mask); But on the other hand, this will make the new API inconsistent with the existing "tep_print_fields()" API bellow.I think we rushed the libtraceevent APIs :-( There's a lot of them I hate, and I agree, passing record would have made more sense. I've been thinking of reworking a lot of them, but we need to add new APIs. tep_print_record_fields() tep_print_record_selected_fields()
OK I can add those two. Thanks! Y.
?? -- Steve
|