[redland-dev] Redland for Revolution & questions

[Dave Beckett]

On Thu, 14 Oct 2004 23:58:26 +1300, rodney tamblyn <rodney at stonebow.otago.ac.nz> wrote:

> Hi everyone,
> 
> Together with a fellow student (Noel Garside) here I've been working on 
> porting Redland to a set of DLLs so they can be called from Runtime 
> Revolution (Rev) programming environment (http://www.runrev.com).  The 
> DLLs include Rev glue and wrapper code allowing callbacks, exchange of 
> variables, and exposing of the Redland API within the Rev scripting 
> environment.   Things have been going pretty well, we've got Redland 
> compiling fine under Windows, including the extra header files and glue 
> files to expose wrapped Redland functions within Revolution.  Once 
> everythings going fine on Windows we'll take a look at doing same on 
> OSX.

I'm not familiar with that environment.

> I'm going to use Redland as part of research project I'm working on to 
> create a collaborative image annotation application for use in medical 
> teaching settings.

Interesting.  The Annotea annotation stuff uses some of the Redland
libraries.
 
> So far we've been able to create a world, storage object (memory) and 
> model, parse in a URI, add statements, and print out a model (using 
> librdf_model_print).
> 
> At the moment I have a problem which I'm hoping list members may be 
> able to help with.  If I create a memory storage I'm able to complete 
> the example above successfully.  However if change the storage params 
> to create a hashes database, and repeat, the persistent database is 
> created successfully on disk, but Rev crashes when calling 
> librdf_model_print.  I've also tried creating a serializer and calling 
> serializer_print_model_to_file with same results.  Executing the same 
> code directly from C as a compiled application (eg example1.exe) runs 
> fine.

So the errors happen when you C standard I/O (printf etc.).  I don't
have an immediate diagnosis since you seem to have it working via C.
Suggests it's the runrev interfering.

> If I create an example which just reads a model from a hashes database 
> on disk, and feed it the database created by Rev, it prints the model 
> out fine.  This shows that there's nothing wrong with the model, or 
> storage, but there's something going on in the serializer when called 
> from Rev.  I think what may be happening is that the serializer is 
> raising an error, but I have not implemented an error callback message 
> (world_set_error).  Perhaps this is then returning control from the 
> external to Rev in such a way as to cause an memory exception error.

That's possible I guess.  The serializer pretty much only calls an
error or warning handler in a few rare occasions such as unable to
write an RDF property in RDF/XML (warning) or some XML
escaping / Unicode encoding problem (error), data corruption (error).

> I'm pretty new to C programming (although very experienced in scripting 
> environments) so I'm hoping someone kind could provide some C examples 
> showing:
> 
> - an example "hello world" redland error handling function - this could 
> for example just print the error to stderr

That's what the default built-in handler does.  You can of course
replace that.

> - code to setup this function (i.e appropriate syntax to configure this 
> via world_set_error)
> - ideally, and example which will cause an error (so that I can test 
> this is working correctly).

That should be pretty straightforward.  examples/example4.c is the
basis of this code, as all it does is serialize.

...later...

I find that the error reporting code in the serializer isn't working.
Because it's so rare, looks like it hasn't been tested recently.

I'll try to make a patch sometime soon.

You can try it in my test code which makes a bad Unicode URI string
which is one way to cause a crash.

> I can then modify the example to send callback messages to Revolution 
> when errors are raised. Once I've got error reporting working for the 
> above, I'll implement functions for serializer_set_error and 
> serializer_set_warning as this is where I think my problem is.
> 
> A couple more questions:
> 1) - if an error occurs and world_set_error has not been set, what 
> happens?

It gets printed to stderr.
> 
> 2) - are warnings and logger actually utilized in the library?  I 
> notice quite a lot of messages being written to stdout and stderr, and 
> just wondered if these shouldn't more appropriately be sent to the 
> logger?

They are used in the library, to report information, warnings and
errors.  The stdout/stderr prints are just the default. Doing nothing
on an error would be a bad default.

> Thanks for help and suggestions in advance.

Dave
-------------- next part --------------
A non-text attachment was scrubbed...
Name: serializer-crash.c
Type: application/octet-stream
Size: 1113 bytes
Desc: not available
Url : http://lists.usefulinc.com/pipermail/redland-dev/attachments/20041015/e9efa5f5/serializer-crash.obj