You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Repro is with Brim 3d473b2 pointing at Zed commit db57e46 which is on the branch associated with brimdata/super#4264.
In a recent group discussion I expressed curiosity as to whether panics currently make their way all the way to the app user. As @mattnibs pointed out, the potential to overflow the stack with a UDF represents one of the easiest ways to trigger a panic, so I gave it a look.
The good news is that the panic is presented. Here's some example Zed to trigger it:
func overflow(n): (
overflow(n+1)
)
from foo
| yield overflow(num)
If the "foo" pool is loaded with a single record {num: 3}, when the program runs we see:
Compare this with the stack from running the same at the CLI:
Comparing the two, I can see a few ways it could be improved:
It would be more readable if there were line breaks like we saw at the CLI.
It might also be more readable in fixed with font.
Most importantly, there's no vertical scrollbar, so the user can't easily see the whole output, such as if they wanted to cut & paste it to send to us for help. Feeling hackish, I realized I could zoom out to make the font super small and then highlight it all, but I suspect many users would not make this connection and instead just cut & paste the parts that were visible.
The text was updated successfully, but these errors were encountered:
Repro is with Brim 3d473b2 pointing at Zed commit
db57e46
which is on the branch associated with brimdata/super#4264.In a recent group discussion I expressed curiosity as to whether panics currently make their way all the way to the app user. As @mattnibs pointed out, the potential to overflow the stack with a UDF represents one of the easiest ways to trigger a panic, so I gave it a look.
The good news is that the panic is presented. Here's some example Zed to trigger it:
If the "foo" pool is loaded with a single record
{num: 3}
, when the program runs we see:Compare this with the stack from running the same at the CLI:
Comparing the two, I can see a few ways it could be improved:
The text was updated successfully, but these errors were encountered: