Add initial .clang-format style and reformat the sources using it

[oerdnj]

This is an initial example of how .clang-format can be configured. It doesn't precisely match the current style, but no tool can do that.

[oerdnj]

If there's something that does specifically bothers you, I can take a look if that's configurable: https://releases.llvm.org/14.0.0/tools/clang/docs/ClangFormatStyleOptions.html

But I prefer the style to stay consistent across all committers over matching every bit of my preferred style.

There's one nit though - I think that even the single line statements should be surrounded by braces - as a defensive programming, I find these:

        if (obj->parsed_uri->model &&
            strncmp(obj->parsed_uri->model, (const char *)token.model, 16) != 0)
            continue;
        if (obj->parsed_uri->manufacturer &&
            strncmp(obj->parsed_uri->manufacturer,
                    (const char *)token.manufacturerID,
                    32) != 0)
            continue;
        if (obj->parsed_uri->token &&
            strncmp(obj->parsed_uri->token, (const char *)token.label, 32) != 0)
            continue;
        if (obj->parsed_uri->serial && strncmp(obj->parsed_uri->serial,
                                               (const char *)token.serialNumber,
                                               16) != 0)
            continue;

extremely hard to parse.

[simo5]

If there's something that does specifically bothers you, I can take a look if that's configurable: https://releases.llvm.org/14.0.0/tools/clang/docs/ClangFormatStyleOptions.html

But I prefer the style to stay consistent across all committers over matching every bit of my preferred style.

Overall I do not mind formatters, but sometimes they make so me silly choices.
Is it possible to make the formatter indent only the changes made to a file but not the entire file in order to preserve exceptions (like so specifically formatted comments) as needed ?

There's one nit though - I think that even the single line statements should be surrounded by braces - as a defensive programming, I find these:

        if (obj->parsed_uri->model &&
            strncmp(obj->parsed_uri->model, (const char *)token.model, 16) != 0)
            continue;
        if (obj->parsed_uri->manufacturer &&
            strncmp(obj->parsed_uri->manufacturer,
                    (const char *)token.manufacturerID,
                    32) != 0)
            continue;
        if (obj->parsed_uri->token &&
            strncmp(obj->parsed_uri->token, (const char *)token.label, 32) != 0)
            continue;
        if (obj->parsed_uri->serial && strncmp(obj->parsed_uri->serial,
                                               (const char *)token.serialNumber,
                                               16) != 0)
            continue;

extremely hard to parse.

I think adding braces here is just fine, I do not mind. In fact I prefer braces even for sinle line if constructs, I just got a bit lazy in this case.

[simo5]

It is not too bad, but let's see if it can be improved (I know subjective, but style often is).

[simo5]

do you mind if I experiment and rebase your branch ?
I think we can get much closer to some of the nits I don't like.

[oerdnj]

I'll also just one shot the single line braces. I thought that clang-format does have support for it already but seems it does not, but I can do that with clang-tidy, but it's not that easy to wire it in, so it's more one-time only.

[simo5]

I'll probably continue review after the weekend, I think we can get close enough.

[oerdnj]

do you mind if I experiment and rebase your branch ?

I think we can get much closer to some of the nits I don't like.

I absolutely don't mind, go ahead. I absolutely also mind a little about the exact style, although going full K&R would be a little hard ;).

Once we nail this, there's also a GitHub Action for clang-format or we can write one.

[simo5]

Not super-happy about how it monkeys with macros, and the whole bizarre things it does with struct initializers with arrays, unless I manually add a comma after the last initializer.

However this comes close enough ...
I am going to try to add a job that checks diffs on pull requests, so that it doesn't complain about intentionally manually arranged stuff.

[simo5]

Ok, this looks reasonably ok.

The Makefile part is not stellar but gets the job done.

What do you think?

[simo5]

Obviously it shows a failure because of the commit where I fixed some clang-format "mistakes".

[oerdnj]

Obviously it shows a failure because of the commit where I fixed some clang-format "mistakes".

There's a command you can put in the comment to enable/disable formatting of the part of the code.

My experience is that it's simply easier to accept that the tool is not perfect and just let it go...

[simo5]

My experience is that it's simply easier to accept that the tool is not perfect and just let it go...

Ok, I guess those macros can live as is ...
When clang will grow some support for better initializer and macro formatting we can revisit this.