OMG I just looked at it. The switch statement at 132 has a return statement at the end of every single case in it instead of a break, and then at the end of the function where one should return, it's left implicit. I'd be embarrassed to have written that.

There's probably worse in there, I was too traumatised to look any further.

2 members found this post helpful.

Rubbish. There is nothing wrong a return statement in place of break in a case clause (if the code is indeed returning), and it would be ridiculous to have a trailing return at the end because it would be unreachable.

You have entirely missed the real issue in the code in question: the case clause fall-throughs. The linux kernel is now fall-through clean. It is likely I would be traumatised by looking at your code (if you actually write any) if that is the standard you go by.

Edit On looking at if further the fall-throughs are OK as they have no code of their own, so I will let you off that one (but not the first point).

3 members found this post helpful.

There's nothing wrong with that. Why bother breaking if you can just return immediately? Some code analysis tools would complain about unreachable code if you had a return at the end of the function. IMHO it's only a problem if it's disjointed, i.e. if you had one case returning and one with a break.

But the biggest problem I see with that file is that it's C and not C++, so instead of RAII you get that 'goto fail' nonsense that belongs in the 90s.

EDIT: Sorry Chris, posted that before I saw your post.

1 members found this post helpful.

Yes, what chrisVV said. Splitting code into separate functions and returning early is the clean, well-behaved, recommended alternative to goto.

There may be nothing wrong with it technically, but it's not good practice. What if one changed the return type of that function? One would have to go through dozens of return statements changing them all. Let it break out of the case and then use a single return at the end of the function and you have just one return to modify. Now, as I said, I'm not a devout follower of the one in, one out, structured programming paradigm and I certainly don't mind using an extra return or so in a guard clause or two when it simplifies the code, but that code is IMO taking it a little too far. If you disagree, then that is your right. No need to get nasty about it.

As for your last sentence, think what you will, I'm not going to rise to that bait.

1 members found this post helpful.

Meanwhile, back at the original question:

1. It doesn't seem to add anything new that I needed done.
2. It replaces a process that a reasonable user-admin could follow and tweek (via scripting) with something opaque to that skill level.

2 members found this post helpful.

Sorry, when you write "OMG I just looked at it ... I'd be embarrassed to have written that" in order to show you are a real slackman complete with six-pack with hairy chest, for code which is absolutely fine in a function returning void (in fact, the best way to write it), not to mention your ridiculous unreachable return statement, then you have to expect to have the position pointed out to you.

Any thread about systemd is bound to end with this kind of "I'm harder than you" drivel.

1 members found this post helpful.

I've no objection to guard clauses, but when you have a switch where every single case ends with a return, then why wouldn't you just put a single return after the switch and break out of the case clauses as normal, especially so when that switch is the last thing in the function and will be followed directly by a return anyway.

If you then have to modify the return type of that function at any time, it's just a single change.

That's a weak self-justification. If you change the return type of the function you will end up rewriting a substantial part of the function logic. In any event, this function is always going to return void. As I say, it is written the correct way.

1 members found this post helpful.

I disagree. It may not be an discussion winning point, but it is a valid benefit of the single return approach, and accusing me of self-justification is just another personal attack on me like your previous posts.

I simply don't like that systemd code, as I said you're free to take an alternative view, but I like my case statements to be terminated with a 'break' even if there's a 'return' right ahead of it, and I like my functions to be terminated by a 'return' rather than falling though to an implicit return because I believe it is good style. The reason I believe this is that IMO doing so reinforces program structure to the reader. The compiler may not care, but I as a human do.

My reaction to that code was over-stated, I'll give you that, but I still stand by my belief that it's an ugly style. Again, you are free to disagree, however I'd ask you to disagree sans the personal attacks in future.

Anyway, no one is interested in this argument, especially me, so I'm done arguing with you here.

1 members found this post helpful.

Ad hominem attacks are inappropriate. Attack the argument, not the person. Your personal attack did not strengthen your position.

4 members found this post helpful.

Just an inoffensive question:

this apply also to the people who derogatory calls a well-known RedHat programmer as "Lennart Voldemort" ?

1 members found this post helpful.

Yes.

2 members found this post helpful.

One can't guarantee that something is inoffensive. Offence is taken, not given. One can say that something is not intended to cause offence.

The contexts are different through. Nobody in this thread is arguing directly with Lennart P [to our knowledge].

4 members found this post helpful.

Unless of course ZhaoLin is LP , all along I thought it was one of Darth's many (possible) alternative accounts.

4 members found this post helpful.