CoralNet sends admin emails whenever spacer jobs return with an error. This has included Deploy jobs where the user-provided image URL was unreachable. That is generally a user-error situation (e.g. didn't provide a public URL), not a spacer bug, so it's not particularly useful for us to get emails when that happens.
In this PR I tried to catch that particular case (without masking other types of errors) and raise a SpacerInputError when it happens. Then when CoralNet's vision backend gets a SpacerInputError back from spacer, it can realize "oh, this isn't a spacer bug, the issue was the inputs given to spacer, so I won't email the admins." I've got a CoralNet-repo PR lined up to complete that logic, though it won't actually work unless this spacer PR is merged into the master branch.
(In the meantime, I saw some assertTrue usages which seemed like they were meant to be assertEqual, so I made a second commit to change those. Let me know if I was mistaken here though.)
CoralNet sends admin emails whenever spacer jobs return with an error. This has included Deploy jobs where the user-provided image URL was unreachable. That is generally a user-error situation (e.g. didn't provide a public URL), not a spacer bug, so it's not particularly useful for us to get emails when that happens.
In this PR I tried to catch that particular case (without masking other types of errors) and raise a
SpacerInputError
when it happens. Then when CoralNet's vision backend gets aSpacerInputError
back from spacer, it can realize "oh, this isn't a spacer bug, the issue was the inputs given to spacer, so I won't email the admins." I've got a CoralNet-repo PR lined up to complete that logic, though it won't actually work unless this spacer PR is merged into the master branch.(In the meantime, I saw some
assertTrue
usages which seemed like they were meant to beassertEqual
, so I made a second commit to change those. Let me know if I was mistaken here though.)