Need help compiling mORMot framework

[shobits1]

I'm trying to test and evaluate mORMot for an upcoming project but when I try to compile `TestSQL3.lpi` using lazarus1.6RC2 the compiler returns the following error:

Code: [Select]

mORMot.pas(22986,7) Error: Internal error 2006111510
Lazarus 1.4.4 wouldn't work since it uses FPC2.6.4 (since the document states FPC2.7.1/3.1.1 )

using Lazarus 1.7 (downloaded from getlazarus.org) the projects compiles ok.

does anyone have insight in the problem.

[JuhaManninen]

I'm trying to test and evaluate mORMot for an upcoming project but when I try to compile `TestSQL3.lpi` using lazarus1.6RC2 the compiler returns the following error:

Code: [Select]

mORMot.pas(22986,7) Error: Internal error 2006111510
Lazarus 1.4.4 wouldn't work since it uses FPC2.6.4 (since the document states FPC2.7.1/3.1.1 )

using Lazarus 1.7 (downloaded from getlazarus.org) the projects compiles ok.

The internal error must be from FPC 3.0. That's what Lazarus 1.6RC2 uses.
I recommend using the version from getlazarus.org if it works. Apparently FPC trunk has fixed the bug that caused an internal error.

[shobits1]

Well, using the FPC trunk it sure compiles the project without errors but running the generated executable (a self test for the entire framework) reports many errors and eventually crashes.

maybe the framework is not so ready for FPC. I'll try and post the problem on mORMot forums and reports back any-finding.

[shobits1]

ok, so A. Bouchez (author of mORMot) response was:

http://synopse.info/forum/viewtopic.php?pid=19447#p19447

Yes, there are issues, mainly due to string constant comparison in FPC.
In Delphi, AnsiStrings with code page do not handle source constant with a particular code page, when compared with another string. It is handled as it code pages do match.

But FPC has a much more pessimistic approach, and depending on the compiler revision, it may - or may not - make a codepage conversion.
A lot of aString='constant' are failing with FPC, whereas aString DO contain the expected 'constant'.
IMHO FPC AnsiString implementation is still broken, and induce a huge performance penalty.
We are waiting for the RTL and compiler to stabilize, until we make some more changes.

In all cases, the interface-RTTI branch is highly recommended.
Perhaps FPC users may help identifying the problems, and help here.

What do you say and what can I do; any help would be appreciated.