CacheOverride StarlingMonkey port #767
Conversation
| @@ -737,16 +737,9 @@ routes.set('/btoa', () => { | |||
| if (error) { return error; } | |||
| error = assertThrows(() => atob("--")) | |||
| if (error) { return error; } | |||
| error = assertThrows(() => atob("__")) | |||
| if (error) { return error; } | |||
There was a problem hiding this comment.
This was fixed upstream in StarlingMonkey, so no longer needs separating.
| @@ -1527,7 +1522,7 @@ routes.set("/backend/timeout", async () => { | |||
| routes.set("/backend/health/called-as-constructor-function", async () => { | |||
| let error = assertThrows(() => { | |||
| new Backend.health() | |||
| }, TypeError, `Backend.health is not a constructor`) | |||
| }, TypeError) | |||
There was a problem hiding this comment.
This is now (intermediate value).health is not a constructor (but still pointing to the line number source of the error in the original source, so the stack retains the context). This is due to the ad-hoc nature of the bytecode name tracing in https://searchfox.org/mozilla-central/source/js/src/vm/BytecodeUtil.cpp#1590, which doesn't carry through to the expression naming for the builtin module. The same thing happens for arbitrary JS expressions though anyway (eg new getObj().health()).
Since new Backend.health() is a pretty extreme edge case, as most JS developers won't attempt to call a function as a class constructor, so I don't think this error message change would be considered breaking.
This ports the cache-override builtin to StarlingMonkey, fixing the TODOs on this subsystem and enabling the tests.
A regression in the ehe error lookup function is also fixed.
Integration Test status: 245 / 726 = 34% passing.