[Pasdoc-main] PasDoc2 (was: public type problem)

[Michalis K. <mic...@gm...>]

(I changed the subject, we diverged from original Ales question here.)

First of all, looks like we already have what we need in trunk. It was
done quickly, without any major redesign of everything, and without
breaking anything that worked, by Arno. That was the solution I was
hoping to see, and we already have it. Remember that my initial response
("patches welcome") to thread "public type problem" was my mistake ---
we already have this implemented in pasdoc trunk. Local types, local
constants etc. already work.

Maybe the current trunk solution needs to be improved --- testcases that
fail are welcome, and let's make a solution that fixes them. If you can
provide examples that fail to be handled by trunk, but are handled by
PasDoc2 --- that would be great.

> 1. Provide a renewed internal structure, that can handle all the new
> syntax elements, and optionally the parsing of the implementation
> section. If you don't like the PasDoc2 model, provide your own or accept
> some other model.

No. Sorry, but that's equivalent to just doing the whole work myself. If
I would have time, I would do this. And I don't want to block other
people's work, but if it breaks pasdoc (introduces regressions and bad
design), I can't merge it to trunk.

To put in bluntly, I'm not interested in PasDoc2 anymore. It introduces
in a complicated and buggy way something that we already have, nice and
bug-free, in trunk.

I tested PasDoc2 again right now (from SVN branches/closed/PasDoc2,
compiled with simple "make", using FPC 2.4.4), and here are the results:

1. It's broken, answering "Fatal Error: No Source Files have been
specified." no matter what is on the command-line. Ok, so I debugged and
fixed it (needs "Generator.Options := Options;" inside TPasDoc.Execute
before calling Generator.Execute).

2. Then I tested it on "a.pas" that was attached to my previous mail in
"public type problem" thread. This tests Ales' example with type inside
class, and tests FPC
http://wiki.lazarus.freepascal.org/class_extensions_examples . PasDoc2
not only fails to parse it (unhandled section in CIO: reserved word
"type"), but also does "Fatal Error: Access violation" at the end.

3. I tried tests/ok_generic.pas from trunk test-suite. PasDoc2 again
fails to parse it, and also again does additional "Fatal Error: Access
violation" at the end.

I'm sorry, but that's the kind of experience I always have with PasDoc2.
New bugs, and no visible new features. The fact that since a long time
you're only talking about PasDoc2, not addressing any bugs that are
mentioned on this list, doesn't help.

So I simply don't see any need to work on integrating the remaining of
PasDoc2.

Michalis