Re: [Swig-devel] Scilab issues

[Ian B. <ian...@gm...>]

Further information on the string handling.

I have a function that takes a std::string, called
get_global_param_string.  I added a debugging command to print out the
length of the string and another string that it should be equal to.  When I
call the function with the argument 'errstring', somehow the string is not
the same length as err.

``` C++
std::string get_global_param_string(std::string ParamName)
{
    std::string err = "errstring";
    std::cout << ParamName.size() << " " << err.size() << std::endl;
...
```
In scilab I do

-->get_global_param_string("errstring")
10 9

This seems to be the root cause of all the problems I have with the scilab
wrapper.  Everything else seems to work fine.

On Mon, Sep 15, 2014 at 2:26 AM, Ian Bell <ian...@gm...> wrote:

> Simon et al.,
>
> A couple of issues with the use of the swig+scilab code that I have run
> into.  I've almost got my code working.  It now compiles at least on
> ubuntu, a major milestone, but doesn't work.  I am not getting any results
> out of my code, but it seems very close.  I guess this is a pretty good
> real-world test case for the wrapper.
>
> OSX is a non starter with my code for the moment.  There are some clang -
> related bugs that I will try to investigate and give an update when I can.
> I haven't tried yet with windows, but I'm not holding my breath.
>
> I use CMake and am trying to integrate into CMake build system.  That has
> its own set of issues, but I think I can work around the CMake problems.
> The scilab problems on the other hand...
>
> builder.sce copies the .cpp files into the TMPDIR directory (yuck.).  This
> is not nice when .cpp files have relative imports of headers that are in
> other folders.  Can you avoid doing that and just keep a list of files to
> be compiled?  To workaround this problem, I had to add all my source
> folders to the list of include folders.  What if you have two files with
> the same name?
>
> I had some problems understanding the values allowed to
> buildverbositylevel, I put in two tiny PR to resolve my confusion, feel
> free to modify.  But that said, even at a verbosity level of 2, I can't see
> what commands are being executed in the makefile itself. Can you please fix
> that?
>
> There is something very funny with string handling.  In C++ land, I have
> std::map<std::string, int>, but for some reason I can't lookup string
> values in the map when wrapping for scilab that I can always do without any
> problems from other wrappers.  In another function, it takes in a string as
> input, and it prints out an error message saying that the exact string I
> passed in is invalid (but it isn't).  Generally it seems string handling is
> broken.  Have you had any experience with this?
>
> To me, it seems like some sort of non-printing character is getting added
> to the strings at the end, or it is getting messed up somehow enough that
> the compare() function doesn't say they are equal, but it's very hard for
> me to say.  In any case, string comparisons with the strings that pass
> through scilab are no longer valid.  I can't say much more than that I
> guess.
>
> Apart from that, seems like things are working, at least kind of.
>
> I will certainly not shed a tear over builder.sce when it goes away.  That
> is such a pain to work with and debug.
>
> Thanks for all your hard work with this!!
>
> Kind Regards,
> Ian
>