RackUnit check-exn fails to catch raised syntax error when error not called directly

[tommymchugh]

Example:

#lang racket
(require rackunit)

(check-exn
   exn:fail:syntax?
   (lambda ()
     (raise-syntax-error #f "Failed" #f #f null "")))

(check-exn
   exn:fail:syntax?
   (lambda ()
     (struct apple (seeds) #:constructor-name pear)
     (apple 53)))

First check passes, but second check never seems to catch that raise-syntax-error (and thus produces an exn:fail:syntax) is called when the identifier is called for a struct that has a different constructor name.

[rfindler]

I believe that's because the exception is being raise at compile time (phase 1) and that's not when check-exn is catching errors.

[sorawee]

I think you will find https://docs.racket-lang.org/syntax/macro-testing.html#%28form._%28%28lib._syntax%2Fmacro-testing..rkt%29._convert-compile-time-error%29%29 useful.

Edit: what about this?

#lang racket

(require rackunit
         syntax/macro-testing)

(check-exn
 #rx"identifier for static struct-type"
 (lambda ()
   (convert-syntax-error
    (let ()
      (struct apple (seeds) #:constructor-name pear)
      (apple 53)))))

[tommymchugh]

Yes that is very helpful thanks a bunch works perfectly! Not that that is a hassle at at all, but for curiosities sake, is there a reason why check-exn does't wrap the convert-compile-time-error procedure around the check-exn thunk procedure? It seems that would be useful especially for syntax errors unit testing which may often produce the error testing for not at runtime.

[rfindler]

The designers of rackunit decided not to do that, IMO, wisely in this case, since catching exceptions is routinely a runtime issue. (Although the extra think doesn't seem necessary to me.) In other words, it is best to pick a phase to evaluate things and not try to do things simultaneously in multiple phases as would be required here. Robby

…

On Thu, Jan 2, 2020 at 9:25 AM Tommy McHugh ***@***.***> wrote: Yes that is very helpful thanks a bunch works perfectly! Not that that is a hassle at at all, but for curiosities sake, is there a reason why check-exn does't wrap the convert-compile-time-error procedure around the check-exn thunk procedure? It seems that would be useful especially for syntax errors unit testing which may often produce the error testing for not at runtime. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2996?email_source=notifications&email_token=AADBNMAPU3SXO7AVJ25XGV3Q3YBPTA5CNFSM4KCDKYY2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEH6S2PY#issuecomment-570240319>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADBNMA4ZKLQVRLAJMHRGILQ3YBPTANCNFSM4KCDKYYQ> .

[tommymchugh]

Gotcha, make sense. Thanks for the assistance on this!

[jackfirth]

A separate check-compile-time-exn form would be very handy though.

[tommymchugh]

@jackfirth I'll build it out and report back. Will reopen for now to see if there is interest in this.

[jackfirth]

@tommymchugh Sounds great 🙂 Heads up: the implementation of rackunit is in the repository racket/rackunit, not racket/racket.

[tommymchugh]

racket/rackunit#114 is the PR with the implementation. Comments, contributions, etc. are welcome!

[sorawee]

Since everything works as expected, can we close this issue, @tommymchugh?

[tommymchugh]

Yup!