Should default value for cstring be nil?

[StefanSalewski]

proc gtk_entry_set_icon_from_icon_name(self: ptr Entry00; iconPos: EntryIconPosition;
    iconName: cstring) {.
    importc, libprag.}

proc setIconFromIconName*(self: Entry; iconPos: EntryIconPosition;
    iconName: cstring = "") =
  gtk_entry_set_icon_from_icon_name(cast[ptr Entry00](self.impl), iconPos, safeStringToCString(iconName))

I would guess yes.

[Dankr4d]

Makes sense. Then you can also get rid of the safeStringToCString call on set prefixed functions. To "clear" it, it's just called without passing the optional cstring, which maybe should be made clear for users in the readme.

[StefanSalewski]

I have done that fix and some more fixes.

You may test with

nimble uninstall gintro
nimble install gintro@#head

This is the upcomming v0.9.4 which I intend to ship this evening.

I have done a test for the gtk3 examples compiled with ARC, and for sdt compiled with ARC. Compiles and runs fine for me. Maybe I will do a few more test later this day.

[StefanSalewski]

We may have to revert this fix due to

nim-lang/Nim#18698

[Dankr4d]

I agree.
Adding more than 1k overloaded functions is maybe not the best, but I don't know how much the compile time would increase.
Also I think we shouldn't use Options, because it bloats the code readability (and is much more to write). Options may be helpful, but I don't think that they're required in this case.
I think the "old way" is currently the best solution, because the user also doesn't need to know anything about c null string pointers.