Deprecate RTLIL::id2cstr()

[rocallahan]

What are the reasons/motivation for this change?

It's exactly the same as log_id(). log_id() occurs 1247 times before this PR, id2cstr() only occurs 77 times, so it makes sense to get rid of id2cstr() to reduce confusion/maintenance work.

Explain how this is achieved.

Trivial change. We leave id2cstr() in place, but deprecated.

[KrystalDelusion]

I prefer id2cstr as the less confusing option given that log_id doesn't log even though it starts with log_*. What's the status of using the IdString natively and not needing to wrap it at all?

[rocallahan]

I prefer id2cstr as the less confusing option given that log_id doesn't log even though it starts with log_*.

Fair point, but on the other hand id2cstr returns something different from IdString::c_str(), so that's also confusing.

The best name here is probably unescape_id(). Which already exists and does what we want. So maybe the best move here is just to replace id2cstr() and log_id() with unescape_id().

What's the status of using the IdString natively and not needing to wrap it at all?

That is already allowed, but it returns the same thing as IdString::c_str(), i.e. the name with the leading \ or $. We need something at the call site to indicate that a leading \ should be dropped. Obvious options:

• A function call (e.g. unescape_id(id))
• A method call (e.g. id.unescape())
• Something in the format string (e.g. log("%#s", id))

Come to think of it, I actually prefer the method call. Let me do a Discourse post about it.