@cadr this reproduces the segfault for me: pk --verbose 4 -t compile-file-pycket.rkt -- -l syntax/location

or just pk -l syntax/location

I found the reason for the segfault, in the bundle we see (name . (srcloc configure-runtime)) which our deserializer doesn’t like (because it assumes the srcloc to have 5 parameters)

I fixed it, but didn’t push it yet because I’m sort of confused to have a srcloc in the name in the bundle, we normally have just a symbol there

ah, i also just found it in the interpreter

that’s a module name

and configure-runtime is a submodule

I think the issue is that deserialize_loop flattens everything into one big conditional

@cadr here’s the log at verbose 8: https://filebin.net/bqhzgr6foqfdp78r

I haven’t found anything wrong yet, though

@samth yeah I have a similar file myself, although I see your latest push is very helpful to keep it under 200MBs

@samth Note that the definitiion of call-with-input-file*61.1 is in a correlated too

yes but that isn’t (obviously) the bug

I tried this:

(module t '#%kernel
  (#%require '#%linklet)
  (define-values (datum->correlated) (hash-ref (primitive-table '#%kernel) 'datum->syntax #f))
  (define-values (s) 'the-symbol)
  (define-values (sexp)
    (list
     'linklet
     '()
     '()
     (datum->correlated
      #f
      (list
      'define-values
      (list (datum->correlated '#f s))
      (list 'lambda '(x) s)))
     (list 'display (list s 1))
     '(newline)))
  (instantiate-linklet (compile-linklet sexp) null))

and it worked

Yeah this works after my fix yesterday