Re: the nul-terminated string helper desk chair rearrangement

[Alexey Dobriyan <adobriyan@gmail.com>]

On Thu, Oct 26, 2023 at 03:59:28PM +0200, Geert Uytterhoeven wrote:
> Hi Steven,
> 
> On Thu, Oct 26, 2023 at 3:52 PM Steven Rostedt <rostedt@goodmis.org> wrote:
> > On Thu, 26 Oct 2023 07:39:44 -0400
> > James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
> >
> > > While it's nice in theory to have everything documented, it's not much
> > > use if no one can actually find the information ...
> >
> > Does kerneldoc provide an automated index? That is, if we had a single file
> > that had every function in the kernel that is documented, with the path to
> > the file that documents it, it would make finding documentation much
> > simpler.
> >
> > Maybe it already does? Which would mean we need a way to find the index too!
> 
> ctags?

ctags is a tool from previous century. It doesn't help that "make tags"
is single-threaded. It needs constant babysitting (loop-like macros,
ignore attibute annotations which masquerade as identifiers). I think
"make tags" became much slower because ignore-list is one giant regexp
which only grows bigger.

> Although "git grep" is faster (assumed you use the "correct" search
> pattern, which can sometimes be challenging, indeed).

I tried QT Creator indexing at some point (which is parallel), it needs
to be told that headers are C not C++. I didn't find a way to tell it
that .c files are C too but F2 jumped to definitions quite well.
Also hovering over identifier/name works (being IDE it understands
popular doc styles).

It can be made to work reasonably well provided that you did "make
allmodconfig" and added few header locations. clangd parses like compiler,
not like human and kernel uses a lot of CONFIG defines so some config
must be chosen.

But I need to recheck all this stuff now that new version was propagated
to distros. It should be better (and less segfaulty :-)