Re: [PATCH 6/9] perf util: Get rid of malloc_or_die() in trace-event-read.c

[Arnaldo Carvalho de Melo <acme@ghostprotocols.net>]

Em Tue, Mar 19, 2013 at 11:13:25PM -0400, Steven Rostedt escreveu:
> On Wed, 2013-03-20 at 12:00 +0900, Namhyung Kim wrote:
> > On Tue, 19 Mar 2013 21:55:02 -0400, Steven Rostedt wrote:
> > > On Wed, 2013-03-20 at 10:14 +0900, Namhyung Kim wrote:
> > >> On Tue, 19 Mar 2013 10:50:02 -0400, Steven Rostedt wrote:
> > >> > On Tue, 2013-03-19 at 17:53 +0900, Namhyung Kim wrote:
> > >> > I think I like the err += func() and check for err < 0, better.

> > >> Okay, I'll change them to err |= func() style if you're fine as Peter
> > >> suggested.

> > > += or |= I'm not picky ;-)

> > Ah, one thing I also care was the short-circuit logic.  I think we don't
> > need to call later functions if one fails, do we?

> Yeah, good point. It still looks ugly, but it does make sense.

Yes, I dislike all this += or |=, it should be normal exception
handling, just like everywhere in the kernel codebase:

	err = foo();
	if (err)
		goto out_err;

	err = bar();
	if (err)
		goto out_foo;

	err = baz();
	if (err)
		goto out_bar;

	err = new_foo();
	if (err)
		goto out_baz;

	return 0;
out_baz:
	baz_cleanup();
out_bar:
	bar_cleanup();
out_foo:
	foo_cleanup();
out_err:
	return err;

----

	That way exception handling code lies at the end of the
function, i.e. in source and binary code it has a lower chance of
polluting brain and CPU caches, and we don't need to call N functions
if we'll bail out when the one of them fails.

	I.e. nothing new here, just follow kernel coding style, move
along :-)

- Arnaldo