[fpc-devel] fpdoc patch - big or small

[Marco van de Voort]

In our previous episode, Graeme Geldenhuys said:
> 
> Michael, there is only one issue I have with fpdoc at the moment. It is
> regarding Marco's recent change regarding resolving links. See the
> following Mantis report for details.
>   http://bugs.freepascal.org/view.php?id=17146
> 
> If I use the code as Marco committed it, then fpGUI's class documentation
> works fine. All links are active and pointing to the correct places. But
> then RTL & FCL's IPF output links are broken.  Swap them around like I did
> in the IPF patch (dglobals.pp line 990-995), then RTL and FCL docs links
> work, but fpGUI's links are broken. :-(

If I understand it well, the code currently tries to first lookup #rtl.xxx
and then #rtl.currentmodule.xxx

The problem was before a relative path could not be an unit name, or be
relative to an unit name (classes.tstringlist)

So the order should only matter for identifiers that duplicate as unitnames.
(and these should then be made with absolute paths, since both  solution then
are not safe)

Anyway, I don't see anything here that could lead to corrupt links.

p.s. 
My agenda for fpdoc in the coming week:
- Make sure that inheritance data is available over package bounderies.
   This will make inheritance trees work over package bounderies (and does
   the prepatory work for interface trees too)

- reduce the noise (warnings). I suspect several of these errors are
 redundant because they are the result of multiple different attempts at
 resolving. This will probably mean a few more service routines that only
 issue a warning if nothing can be found using multiple means. But that is
 only a hunch, I still have to read up on this part of fpdoc.