Consider getting rid of `assert` calls in compiler that could be user-facing

[fhammerschmidt]

This yields cryptic errors where the user does not really know what's going on.

For instance in bsb_pkg_types.ml if the package name begins with an @ it asserts that there is a slash somewhere afterwards. If there is no /, you get:

Fatal error: exception File “bsb_pkg_types.ml” error

Related forum post: https://forum.rescript-lang.org/t/getting-fatal-error-exception-file-bsb-pkg-types-ml-no-solution-found/5091/2

[zth]

I agree. There's a plethora of asserts in the compiler though. But fixing them continuously would indeed be great.

[cometkim]

I'm facing #6539 and #6585

I think there are two cases:

• Lazy error handling: This can be solved by simply providing an explicit error code and an appropriate description of the situation.
• Wrong assumption: The compiler reached an invariant case never expected. Ideally, if the assumption isn't wrong, we can fix this by modifying the code, but it's not always simple. At a minimum, we should provide a message template to guide users to recognize and report this as a compiler bug. If it is absolutely necessary, there should be enough comments in the codebase to explain the assumptions that cannot be violated.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[fhammerschmidt]

not stale