Peter Hill

This turns out to be a common extension, and was never standardised. Do you know if it's actually used anywhere?

Ah, this is specifically in `character_length`, it just needs `optional(',')`

I've had a stab at this and I think it's going to be _very_ complicated because it clashes with variable attributes. I've found 12 instances in 40M lines of code...

This is an unimplemented extension, should be possible to add though. There might be some fun clashes with `variable_modification` though.

Here's what it should parse as: ```lisp (translation_unit (program (program_statement (name)) (variable_declaration type: (intrinsic_type) attribute: (type_qualifier (argument_list (number_literal) (number_literal))) declarator: (identifier)) (do_loop_statement (loop_control_expression (identifier) (number_literal) (number_literal)) (assignment_statement left: (call_expression (identifier)...

From a quick look, it seems most places are using it in the sense of the comment (that is, checking if there's a boundary, and if so, use one-sided differences),...

Ah, that's a pain! Thanks for the heads up. PRs very welcome :)

It's possible to deprecate the macro itself, here's one way to do it ([in action](https://godbolt.org/z/aqWq7f7Kf)): ```c #ifdef _MSC_VER #define _FillValue _Pragma ("message( \"'_FillValue' macro is deprecated, please use NC_FillValue\")") "_FillValue"...

Seems like a reasonable attempt! That whole bit is getting quite long now, but be nice to pull it out into a `configure_bout_config_script` function somewhere else to keep the main...

If we think this is not super useful, should we just delete it?