igraph_error_t should be an enum

[szhorvat]

igraph_error_t should be an enum, now it's a typedef for an int.

Be sure to test with C++.

[GroteGnoom]

Be sure to test with C++.

It should just pass all our tests, right? Or do you think extra tests are required?

[szhorvat]

Thinking about it, I think the C++ parts of igraph itself should be sufficient to test it. If the library compiles, we're fine.

[GroteGnoom]

igraph_warning takes an igraph_error_t as one of its arguments, but that argument is 0 in one place in the source code, and -1 everywhere else (like in IGRAPH_WARNING). -1 is of course not allowed if igraph_error_t is an enum. Should we just remove the parameter?

[szhorvat]

I was always of the opinion that this parameter should be removed.

Ping @ntamas @vtraag for input.

[ntamas]

Yes, this can be removed (in igraph 0.10). Please either do it in the develop branch now or add it to the list of breaking changes in the wiki so we don't forget it.

[GroteGnoom]

Fixed in #1948