[fpc-devel] Hi listers, , , , I am trying to translate a C headerwith H2PAS

Hans-Peter Diettrich DrDiettrich1 at aol.com
Tue Jun 21 12:35:15 CEST 2011


Skybuck Flying schrieb:

> What's so hard about converting a C to a Pascal file, even partial 
> conversion might still be somewhat usefull.

Many tools (e.g. h2pas) exist for partial conversions (of header files). 
A reliable conversion requires e.g. the exact header files for all data 
types. Remember that C has no fixed-size data types at all?

> In can imagine the converter might need to know where certain "includes" 
> are but that's pretty much it...

The header files in turn rely on certain compiler-supplied #defines, 
which also must be supplied. Fortunately the number of existing 
compilers is not so big, so that I started to supply some configs for 
gcc, BCB and VS. The user still has to select the applicable compiler, 
or derive his own default settings from the supplied samples.

Also the interpretation of the source code, even the C language specs, 
varies amongst compilers. I had to add several hacks for e.g. MSVC, and 
specific #pragma implementations may be required for other compilers. 
This means that the user *must* select an existing compiler, what should 
be not so hard in the concrete case (MSVC), but may lead to 
unpredictable mismatches in data types in other cases (gcc), which can 
make the result incompatible with concrete (binary) libraries.

> Some gui where additional files/paths could be set would be sufficient ?!?

That's doable for directories, of course, but I postponed any GUI 
development until solutions for the more important problems can be 
found. Most critical is the determination, which #defines should 
translate into named constants, subroutine calls, or must be substituted 
by the preprocessor. The Windows headers with their *thousands* of 
#defines still wait for an according classification - be manual or 
automatic. This was the point when I stopped further work on ToPas. It 
works as-is, but its correct usage (for reliable results) requires some 
additional knowledge.

I do *not* want to provide unreliable software, that is easy to use, but 
may provide incorrect results when configured in a wrong way. While FPC 
or Lazarus report errors when used with a wrong configuration, a source 
code converter cannot determine whether its configuration is okay, 
except for missing or definitely incompatible header files. Even then 
it's up to the user to find the *real* problem, and to adjust his config 
accordingly. He who can find out the trouble maker will also be capable 
of correting the config accordingly, be with or without a GUI.

If you don't like my tools, then either improve them or don't use them. 
I'm open to constructive criticism, but the implementation of your 
suggestions is not as simple as you think. Instead you should understand 
the real problems first, and how they could be solved in concrete code.

DoDi




More information about the fpc-devel mailing list