Use restart in test_code()
#2153
Conversation
| @@ -175,7 +167,7 @@ test_code <- function(test, code, env, reporter, skip_on_empty = TRUE) { | |||
|
|
|||
| debug_end <- if (inherits(e, "skip_empty")) -1 else 2 | |||
| register_expectation(e, debug_end) | |||
| signalCondition(e) | |||
| invokeRestart("end_test") | |||
There was a problem hiding this comment.
This now seems like a clear bug — resignalling a condition inside a calling handler would have just duplicated the condition.
| } | ||
| handle_fatal <- function(e) { | ||
| handled <<- TRUE | ||
| # Error caught in handle_error() has precedence |
There was a problem hiding this comment.
No longer needed because we jump out of both the tryCatch + withCallingHandlers stack with a restart.
There was a problem hiding this comment.
hmm... According to the comment on line 123, register_expectation() might fail, and one of the reasons the test_error setup was done is so the original error, rather than the internal one, would be registered here.
But we call register_expectation() again below (albeit with a different debug_end) so I'm not sure why this really matters.
Perhaps just remove the comment on line 123?
| } | ||
| handle_fatal <- function(e) { | ||
| handled <<- TRUE | ||
| # Error caught in handle_error() has precedence |
There was a problem hiding this comment.
hmm... According to the comment on line 123, register_expectation() might fail, and one of the reasons the test_error setup was done is so the original error, rather than the internal one, would be registered here.
But we call register_expectation() again below (albeit with a different debug_end) so I'm not sure why this really matters.
Perhaps just remove the comment on line 123?
No description provided.