-
Notifications
You must be signed in to change notification settings - Fork 824
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat(interface-types) Greatly improve errors #1285
feat(interface-types) Greatly improve errors #1285
Conversation
bors try |
tryBuild succeeded |
The new `errors` module contains structure to represent errors, instead of using basic strings. The first usage is in the interpreter itself.
04bf6d2
to
4ffb158
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me other than my comments!
"`{}` failed because there is not enough data on the stack (needs 2).", | ||
instruction_name, | ||
)) | ||
if memory_view.len() < pointer + length { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this may be affected by the same bug I recently fixed in WasmPtr,
Given a String, "hi" at address 0xFFFE the memory looks like
0xFFFE: 'h'
0xFFFF: 'i'
ptr (0xFFFE) + len (2) = 0x10000, failing this bounds check. The fix, in general, is annoying when you handle length = 0 and zero sized types.
See #1272 for my fix.
Better to have false positives than false negatives in bounds-checking code, but leaving a comment here as a heads up. It might make sense to do a full pass for these types of bugs and fix them in a separate PR if you think it might clutter this one.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To clarify more, we want to make sure out of bounds pointers are never derefed, so ptr = 0x10000 and length = 0 will not be caught by the naive solution of changing <
to <=
here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not sure I understand. I've pushed a new commit to use inclusive an range, and consequently a special condition to handle empty string, but I'm not sure there is a bug right now. The test_memory_to_string
already reads at bounds.
bors r+ |
Build succeeded |
This PR is build on top of #1284. It must be merged first. View the exact diff.Errors in
wasmer-interface-types
were just&'static str
, which isn't quite great 😉. This PR introduces theerrors
module, with structures that represent errors. ADisplay
implementation maps the errors to strings.