Write test repro for "We assume an internally valid datamodel before mutating." from example schema

[tomhoule]

Command: prisma introspect
Version: 2.17.0
Binary Version: 3c463ebd78b1d21d8fdacdd27899e280cf686223
Report: https://prisma-errors.netlify.app/report/13029
OS: x64 darwin 19.2.0

JS Stacktrace:

Error: [libs/datamodel/connectors/dml/src/model.rs:166:14] We assume an internally valid datamodel before mutating.
    at ChildProcess.child.on (/Users/mac/Desktop/codestates/insta-clone/node_modules/prisma/build/index.js:39181:28)
    at ChildProcess.emit (events.js:182:13)
    at ChildProcess.EventEmitter.emit (domain.js:442:20)
    at Process.ChildProcess._handle.onexit (internal/child_process.js:240:12)

Rust Stacktrace:

   0: backtrace::backtrace::trace
   1: backtrace::capture::Backtrace::new
   2: user_facing_errors::Error::new_in_panic_hook
   3: user_facing_errors::panic_hook::set_panic_hook::{{closure}}
   4: std::panicking::rust_panic_with_hook
   5: std::panicking::begin_panic_handler::{{closure}}
   6: std::sys_common::backtrace::__rust_end_short_backtrace
   7: _rust_begin_unwind
   8: core::panicking::panic_fmt
   9: core::option::expect_failed
  10: dml::datamodel::Datamodel::find_relation_field_mut
  11: sql_introspection_connector::re_introspection::enrich
  12: sql_introspection_connector::calculate_datamodel::calculate_datamodel
  13: <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
  14: <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
  15: <futures_util::compat::compat03as01::Compat<Fut> as futures::future::Future>::poll
  16: <futures::future::either::Either<A,B> as futures::future::Future>::poll
  17: <futures::future::then::Then<A,B,F> as futures::future::Future>::poll
  18: <futures::future::either::Either<A,B> as futures::future::Future>::poll
  19: futures::task_impl::Spawn<T>::enter::{{closure}}
  20: <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
  21: introspection_engine::main
  22: std::sys_common::backtrace::__rust_begin_short_backtrace
  23: std::rt::lang_start::{{closure}}
  24: std::rt::lang_start_internal
  25: std::rt::lang_start

[tomhoule]

Error reports: 13029 and 13036

[tomhoule]

And 13045

[do4gr]

They all seem to be triggered during the re-introspection part of introspection. This could be fixed by our changes to datamodel parsing, but not sure.

It's hard to reproduce at the moment since we only have one of the inputs to replay introspection in the correct shape in the error reports (the datamodel). The second one, the introspected sql schema we only have as a string, but have no easy way of deserializing it into a sql schema again.

[janpio]

Do we have an easy way to get from the schema/datamodel we have, to the sql schema string version of that? Then we could diff the both string representations - which should be pretty close - and see where they are different. This could possibly lead to us being able to reverse engineer the missing SQL - or come up with a theory what is happening directly.

[do4gr]

We'll reintroduce deserializable logging that should allow us to replay the issues. I manually did that for 13029, but it's too much monkey work to want to do it again.

[do4gr]

Unfortunately we were not collecting the information in our error reporting in a format that allows us to replay the bug. That has changed with 2.19 which should be out next Tuesday.

I'm closing this one now because we cannot act on it.