Hi,

I am writing a Freepascal IDE in Linux+GTK2 (using freepascal, of course!). I attach a screenshot for your curiosity, just in case.

At this time I am trying to finish a good auto-completion, so I need to extract symbols from pre-compiled PPUs (especially the runtime standard libraries). Right now I use ppudump(1) to extract symbols, but I could switch in future to the use of ppu.pas. I noticed that different freepascal releases do not share a common PPU file format, so using ppu.pas inside the IDE would force the user to use the same compiler version as the one used to compile the IDE. On the other hand, I noticed that the release 2.4.x is the first one which reports "v100" for a PPU file, which hopefully indicates that future releases could be PPU-compatible...

Now, I would ask an advice: is it better to stick to ppudump, at a price of some reduction in speed, and the risk that its output could change in future, invalidating the IDE parsing routines? Or is it better to use ppu.pas in the IDE, having confidence that ppu.pas will parse PPU file generated by different versions/releases of freepascal?

Thanks in advance for your reply, and best regards.

Doriano Blengino

Well, as far as I know, ppu simply skips blocks that it does not know, so I would think it is safe to use.

Michael.

First of all: please mention in the future that you are not subscribed (also in replies, because people may forget), so that people who reply know that they should CC you.

The "version number" in PPU files is just an increasing byte. 100 is not "1.00", it's "one hundred". It has no special meaning.

Now, I would ask an advice: is it better to stick to ppudump, at a price of some reduction in speed, and the risk that its output could change in future, invalidating the IDE parsing routines?

You should use ppudump.

Jonas

Op 2010-06-22 14:40, Doriano Blengino het geskryf:

At this time I am trying to finish a good auto-completion, so I need to extract symbols from pre-compiled PPUs (especially the runtime standard libraries). Right now I use ppudump(1) to extract symbols, but I could

Interesting, is that actually faster/better than parsing the sources?

Regards,
- Graeme -

Yes. Additionally, it also works if no sources are available.

Michael.

...and also deserve no knowledge about the compiler settings at compile time.

DoDi

Jonas wrote

First of all: please mention in the future that you are not subscribed (also in replies, because people may forget), so that people who reply know that they should CC you.

Sorry - I didn't notice (is it mentioned somewhere?).

** I'am not subscribed to this list, at present **

I made some tries with two or three different releases, making ppudump parse "its own" ppu files and ppu files from other releases. At different extents, different problems arose. In one bad case, ppudump (version about 2.2.0) crashed while parsing ppu files from its own release, trying to allocate an enourmous block of memory. A self compiled copy, took from 2.4.0 sources, did the same. Many times it complained about "!! has more information" (who cares), "!! refid out of range" or similar (critical). Anyway, using external ppudump, seems to be the safest way to go... the only problem is that ppudump takes binary informations and converts them in a human-readable form, which are later parsed and converted again in binary form... it takes several seconds to parse the big gtk2.ppu file.

Regards,
Doriano

I think it's normal to expect that people replying to a message on a mailing list may only reply to the mailing list if they don't know that you are not subscribed.

> Well, as far as I know, ppu simply skips blocks that it does not > know, > so I would think it is safe to use.

First of all, it is not safe. While ppudump indeed skips blocks that it doesn't know, the contents of the blocks can also change between different versions.

ppudump is indeed often out of date because it's seldom used by compiler developers and not automatically tested.

Jonas