Lars Schneider <larsxschneider@xxxxxxxxx> writes:
> OK. Something like the patch below would work nicely.
Yeah, something along that line; it would eliminate the need to
worry about a field named "stdin" ;-)
But ...
> while (children_to_clean) {
> struct child_to_clean *p = children_to_clean;
> children_to_clean = p->next;
> +
> + if (p->clean_on_exit_handler) {
> + p->clean_on_exit_handler(p->pid, in_signal);
> + }
... the application that used run_command() API would want to be
able to pass extra piece of data that is appliation-specific for the
child being killed, so it may make sense to extend the function
signature to take a pointer to "struct child_process" for the child
process being killed, together with a new field added to "struct
child_process" that is "void *exit_handler_cbdata;", perhaps?