[v3,media] Use common error handling code in 19 functions

Commit Message

Mauro Carvalho Chehab May 4, 2018, 5:50 p.m. UTC

Em Fri, 4 May 2018 18:08:59 +0200
SF Markus Elfring <[e-mail]> escreveu:

> > Adjust jump targets so that a bit of exception handling can be better
> > reused at the end of these functions.  
> 
> Why was this update suggestion rejected once more a moment ago?
> 
> https://patchwork.linuxtv.org/patch/47827/
> lkml.kernel.org/r/<[e-mail]>
> https://lkml.org/lkml/2018/3/9/823

Taking just the first diff there as an example:


and probably even produce the same code, as gcc will likely optimize
it. It is also easier to review, as the error handling is closer
to the code. On the other hand, there's nothing wrong on taking
the approach you're proposing.

In the end, using goto or not on error handling like the above is 
a matter of personal taste - and taste changes with time and with
developer. I really don't have time to keep reviewing patches that
are just churning the code just due to someone's personal taste.

I'm pretty sure if I start accepting things like that, someone
else would be on some future doing patches just reverting it,
and I would be likely having to apply them too.

So, except if the patch is really fixing something - e.g. a broken
error handling code, I'll just ignore such patches and mark as
rejected without further notice/comments from now on.


Thanks,
Mauro

Comments

SF Markus Elfring May 5, 2018, 7:53 a.m. UTC | #1

> @@ -656,18 +656,18 @@  static int dvb_dmxdev_start_feed(struct dmxdev *dmxdev,
>  	tsfeed->priv = filter;
>  
>  	ret = tsfeed->set(tsfeed, feed->pid, ts_type, ts_pes, timeout);
> -	if (ret < 0) {
> -		dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
> -		return ret;
> -	}
> +	if (ret < 0)
> +		goto release_feed;
>  
>  	ret = tsfeed->start_filtering(tsfeed);
> -	if (ret < 0) {
> -		dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
> -		return ret;
> -	}
> +	if (ret < 0)
> +		goto release_feed;
>  
>  	return 0;
> +
> +release_feed:
> +	dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
> +	return ret;
>  }
> 
> There's *nothing* wrong with the above. It works fine,

I can agree to this view in principle according to the required control flow.


> avoids goto

How does this wording fit to information from the section
“7) Centralized exiting of functions” in the document “coding-style.rst”?


> and probably even produce the same code, as gcc will likely optimize it.

Would you like to clarify the current situation around supported
software optimisations any more?


> It is also easier to review, as the error handling is closer
> to the code.

Do we stumble on different coding style preferences once more?


> On the other hand, there's nothing wrong on taking the approach
> you're proposing.

Thanks for another bit of positive feedback.


> In the end, using goto or not on error handling like the above is 
> a matter of personal taste - and taste changes with time

Do Linux guidelines need any adjustments?


> and with developer. I really don't have time to keep reviewing patches
> that are just churning the code just due to someone's personal taste.

I tried to apply another general source code transformation pattern.


> I'm pretty sure if I start accepting things like that,
> someone else would be on some future doing patches just reverting it,
> and I would be likely having to apply them too.

Why?

I hope also that the source code can be kept consistent to some degree.


> So, except if the patch is really fixing something - e.g. a broken
> error handling code, I'll just ignore such patches and mark as
> rejected without further notice/comments from now on.

I would find such a communication style questionable.
Do you distinguish between bug fixes and possible corrections for
other error categories (or software weaknesses)?

Regards,
Markus

Patch

diff --git a/drivers/media/dvb-core/dmxdev.c b/drivers/media/dvb-core/dmxdev.c
index 61a750fae465..17d05b05fa9d 100644
--- a/drivers/media/dvb-core/dmxdev.c
+++ b/drivers/media/dvb-core/dmxdev.c
@@ -656,18 +656,18 @@   static int dvb_dmxdev_start_feed(struct dmxdev *dmxdev,
 	tsfeed->priv = filter;
 
 	ret = tsfeed->set(tsfeed, feed->pid, ts_type, ts_pes, timeout);
-	if (ret < 0) {
-		dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
-		return ret;
-	}
+	if (ret < 0)
+		goto release_feed;
 
 	ret = tsfeed->start_filtering(tsfeed);
-	if (ret < 0) {
-		dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
-		return ret;
-	}
+	if (ret < 0)
+		goto release_feed;
 
 	return 0;
+
+release_feed:
+	dmxdev->demux->release_ts_feed(dmxdev->demux, tsfeed);
+	return ret;
 }

There's *nothing* wrong with the above. It works fine, avoids goto