Re: [Widelands-public] Sourcecode beautification

[Florian W. <Flo...@pf...>]

Am Donnerstag, 1. Januar 1970 00:59 schrieb Philipp Niemann:
> |	static Building_Descr* create_from_dir(Tribe_Descr* tribe,
> |-		const char* directory, const EncodeData* encdata);
> |+	                                       const char* directory, const

> OLD: Two tabs in front of line continued
> NEW: ONE tab and lots of spaces in front of line continued
> Seems like astyle wants the open parenthesis to be continued on the same
> column in the next line.

This is indeed what is happening. And I think it's the right thing to do, f=
rom=20
the viewpoint of indentation and disregarding line length.

> time to time I do use 80 char width terminals, the above example is a
> strong "no" to automatic use of astyle.

80 characters is, another formatting problem we currently face and I agree =
it=20
has to be addressed - preferrably at the same time as indentation. (cf. las=
t=20
paragraph for software 'solutions')

Two "but"s:
1. Notice that, in your example, the un-astyled line is <80characters only=
=20
because the author manually made it so. An already reindented line can be=20
treated the same way, once it is found.

2. There already is a measure of code that has this problem (ex.: bob.h::vo=
id=20
start_task_movepath).

Addressing #1:
Provided one does use astyle and breaks annoying lines manually, both party=
s=20
could probably be happy:
* people with big screens will have relatively big indents (say, 8sp).=20
   Nevertheless, they break the lines at 80characters. Any smaller indent t=
han=20
   theirs (say, 3sp) will display correctly, only a few characters at the e=
nd=20
   don't get used
* the other way around, nobody cares. Big screens are big, even 130char lin=
es
   are no problem.

So reindenting is "only" a matter of introducing additional long lines;=20
probably not that many. I'm not saying we shouldn't care - but it's not muc=
h=20
additional trouble. (I can't quantify that yet, will test tomorrow)

> indentation handling. But I hate programs messing up my thoughtfully
> written things.

=2E.. and, in this case, doing so because you yourself ordered it. ;-)

> What bothers me is that I miss the meaning of a patch for hundreds of
> lines of indentation changes or trailing spaces removement. But still I

My opinion is that reformatting patches should contain nothing else and vic=
e=20
versa.

Yes, I'm painfully aware that I've acted differently time and time again. :=
=2D(
Mostly because of "Oops, my editor lost that setting _again_. But I'm too l=
azy=20
to clean the patch, I want it out _now_ and go to bed." =20

> About to read its manpage.

I couldn't find a way to solve the problem. The only program that does fold=
ing=20
for long lines is gnu indent (and some very expensive commercial ones). But=
=20
indent doesn't understand C++. So it's either spend money or fold manually.

To lessen the impact, one might:

1. use --max-instatement-indent=3D5. This way the additional indentation wo=
uld=20
    be 5 spaces max. (or any other n)

2. use --force-indent=3Dtab=3D8 instead of --indent=3Dtab=3D3. This uses as=
 many tabs=20
    as possible (assuming the given width) to still get the same indentatio=
n=20
    as with spaces. If your local tabwidth is less, stuff moves to the left=
=20
    much more.

#1 is crap, IMHO. I suggest you try #2 and see how it looks.

We might also find a way to first break long lines with indent, then do a n=
ice=20
formatting with astyle. This will take a lot of fiddling with indent=20
parameters. The downside: it will still very likely touch the sourcecode in=
=20
ways we won't like. There is no way to make indent _not_ change braces and=
=20
such (and barf immensely at some C++ constructs). astyle, even if we found=
=20
the perfect parameters, probably couldn't undo the damage indent did.

I definitely wouldn't like that solution, and I guess I'm not alone with th=
e=20
sentiment.

>cover some files
>step by step beforehand to get an idea of the quirks of astyle. As from

The current state of sound_handler.* is:
write-->astyle-->manually break to 80chars-->astyle.
=46inal indentation was done with the parameters you used in your trial.

So have a first look to see what it could be like. :-)

=2D-=20
Mit freundlichen Gr=FC=DFen,
=46lorian Weber

PGP key id: A34C32F9