scoder

Argh, right. I forgot about the runtime configuration part. It's incredible in retrospect how much work we put into this feature. It certainly *is* a helpful debugging feature, but it...

~BTW, I also think it's fine to change the default in `options.c_line_in_traceback` to compiling this out. If users really want this feature, it's easy to enable with CFLAGS at compile...

Right. That gives us this then: ```python # Error handling and position macros. # Using "(void)cname" to prevent "unused" warnings. mark_errpos_code = ( "#define __PYX_MARK_ERR_POS(f_index, lineno) {" f" {Naming.filename_cname} =...

>* 0 for disabled >* 1 for enabled >* -1 for runtime My first thought was -1, too, but I think 2 is better. It's like 1 (i.e. on) but...

> mention versions We should still make it clear that the macro also works in older versions than 3.1. Users should generally migrate away from the options, not just for...

> default behaviour should be equal to setting macro CYTHON_CLINE_IN_TRACEBACK=0. I agree with this. Let's see if we can find a way to make it easy for users to use...

I've proposed such a 2-macros implementation, making some handwavy assumptions. - Most users do not need C lines in tracebacks for their user facing production releases and prefer smaller modules....

I don't think it needs to be part of the shared library. Both the .c file and the .so file are generated, so we can continue to compare their time...

> it has to be a warning to avoid breaking existing working code. Yes. Let's make it a visible warning so that users can adapt, and make it clear in...

This can be done with the `AutoDocTransform`, which we use to prepend signatures to docstrings. We could use the same string as `__text_signature__`. Does this only apply to `__init__`, BTW?