Le 24/07/2022 à 14:26, Jean Abou Samra a écrit :
Hi, For context, I'm trying to fix issues with source locations from embedded Scheme code in LilyPond. Essentially, with Guile 2.2, errors are displayed without any source location info because the evaluator doesn't provide them. The compiler does, so I'm investigating whether we can use it. So far, I've been experimenting in the REPL, and tearing my hair because I couldn't get locations to work. My test started to become inconvenient to modify in the REPL at some point, so I switched to putting that code in a .scm file. To my great surprise, that made locations suddenly work. Example: (use-modules (system base compile)) (define code-str " (begin (debug-disable 'backtrace) (define (func1 x) (func2 x) (func2 x) (func2 x)) (define (func2 x) (display (func3 x)) (write (func3 x)) (map write (func3 x))) (define (func3 x) (error \"Oops\")) (func1 #f)) ") (define code-synt (call-with-input-string code-str (lambda (port) (set-port-filename! port "fake.scm") (read-syntax port)))) (compile code-synt) When I paste this in the REPL, the given error is ice-9/boot-9.scm:1685:16: In procedure raise-exception: Oops When I put it in tests.scm and run "guile3.0 tests.scm", the error is fake.scm:15:4: In procedure func1: Oops When I put it in a LilyPond file, the result is Oops (without any location info at all). Does anyone have a clue about what causes these differences? Is it expected? Should I report it as a bug? Also, is there a workaround?
Well, 10 minutes later, I have my answer. Both the REPL and LilyPond use catch to be robust against errors, printing the backtraces but not exiting themselves. I need to investigate if there is a way to get the location when catching an exception. Sorry for the noise, Jean
