build.rkt Breaks When Using Racket 7.0 about curr

[schanzer]

Paging @ds26gte - can you look into this?

[ds26gte]

@schanzer, I'll look into it, but as a reminder, this is the same problem that I ran into for which I asked for logs from you and @kfisler around Aug 15, since neither of you were hit by the bug. Looking at the latter's log, I noted that she was using Racket v6.12, not v7 as I was.

There seems to have been a change in how syntax-parameters (~= symbols as macros) work between Racket v6.12 and v7, but I couldn't find it documented. Our use of the syntax-parameter current-the-unit-description seems to trip this somehow (i.e., it's set to something useful in v6.12, but #f in v7). Maybe because this is a syntax phase issue, instrumenting when something is happening didn't help at all at the time, hence my decision to downgrade to v6.12 for the time being.

[blerner]

@ds26gte I pinged folks on the Racket slack, and Sam mentioned racket/racket#2035 and confirmed that he's pretty sure syntax-parameters did change in 7.0

[blerner]

Further follow-up from Matthew: "Offhand, it looks like a problem with syntax transformers and set! – probably a bug in the expander, although it’s possibly in syntax parameters. The expanded code has (#%app the-unit-description the-unit-description ....), and I think it was supposed to be (set! the-unit-description ....)."

Probably worth following up with him offline from this bug, until we know what's going on and how to fix it.

[mflatt]

Yes, this seems to be a bug in syntax-parameterize. Specifically, it looks like a missing (require (for-template '#%kernel)) in “stxparamkey.rkt” to make an identifier-binding comparsion on set! work as intended. I should be able to push a repair (for the next release) later today.

[schanzer]

Thanks @mflatt and @blerner !!

[schanzer]

@doshiro is this still an issue for you?