Bugs/issues after the big destructuring PR #305
Comments
…or spread elements in call arguments. Fixed various issues related to destructuring assignments. See #305.
|
Thanks for reporting these. I think I have fixed most of them, except for the patterns in catch, shall have a look later. |
|
Thanks very much for the extremely fast fixes 🙇 I did rerun my tests and apart from the try/catch stuff all seems to now work. Additionally, I now noticed that tests such as language/eval-code/direct/func-decl-a-following-parameter-is-named-arguments-declare-arguments-and-assign.js are now failing due to not throwing SyntaxError where it's expected instead of the previous error ("arguments is a reserved word in strict mode") that shows me, that we(k6) probably shouldn't run this test in k6 as it's While I would think this particular (or all the other tests failing around this) aren't really practical in themself this seems to be some kind of evaluation ordering bug(?) 🤔 so it seems interesting from that perspective. |
|
I believe all issues mentioned here are now fixed. If anything else crops up please open separate issues. |
* Avoid calling iterator's return method again when it throws * Fixed dynamic variable resolution when a parent lexical binding exists * Support dereferencing in ExportTo(). Closes dop251#300 * Removed the mention of the es6 branch, updated the section about new features development. * Destructuring assignments, rest and spread properties, default function parameters (dop251#303) * Fixed argument variable reference resolution in stashless functions * Fixed binding order for global function declarations when there are duplicate names. Fixes dop251#344. * Make sure variables dynamically bound in parameter scope conflict with parameter bindings. See dop251#305. * Fixed setting the stack pointer in enterFuncBody. Fixes dop251#309 * Support for assignment patterns in for-in and for-of loops. Support for spread elements in call arguments. Fixed various issues related to destructuring assignments. See dop251#305.
…formats, ... (#47) * Fixed typed arrays' defineProperty and indexing. Fixes dop251#308. * Fix Proxy creation panicing on target not callable * Do not modify sb for variadic calls as it breaks local variables resolution. Instead place a marker on stack and use it to count the number of args. Fixes dop251#311. * Support patterns in catch clause. See dop251#305 * Fixed the handling of Symbol properties in destructuring assignments. Fixes dop251#312. * Ensure ToPropertyKey happens earlier when assigning computed keys. Fixes dop251#312. * Aligned detached buffer semantics with the latest specs. Fixes dop251#315. * More typed arrays fixes * Fixed accessor and property key function names. Fixes dop251#314. * Fixed possible panic when sorting non-standard arrays. * Added nil-safety checks for values returned by get*() methods * Fixed panics in parser on some invalid inputs. Fixes dop251#318. * Arrow function (dop251#319) * Implemented arrow functions. See dop251#304. * Define the name property for anonymous functions (including arrow functions) * Treat "arguments" as a lexical binding to match the latest specification * Allow arrow functions to contain 'use strict' for simple parameter lists. Fixes dop251#323. * Fixed argument variable reference resolution in stashless functions * Treat date-only formats as UTC and date-time as local timezone. Added support for additional datetime formats. Fixes dop251#281, fixes dop251#292. * Use correct createArgsRest variant when arguments are not in stash. Fixes dop251#327 * Fixed formatting for go 1.17 * Report 'length' as own property for Go slices. Fixes dop251#328.
First great job on implementing all of that functionality 👏 🙇 🎉 and thank you very much.
As I have previously mentioned k6 runs the majority (no async/await, promise stuff) of the tc39 testsuite including tests that fail (for various reasons) to test the goja+babel setup we have and its problems. As such we have some new tests that are now failing without disabling plugins (and even more that are now passing 👏 , unfortunately, I am here to only tell you about the failing ones, sorry :( ):
({}) = 1;should be a syntax errorDuplicate declaration nruntime error: invalid memory address or nil pointer dereferenceas well as it's ...lex-open.js test and the same two tests forfor-inThen of course I started disabling babel plugins of now implemented stuff ;).
Disabling the
transform-es2015-destructuringandtransform-es2015-computed-propertiesplugins of babel also shows that:for-ofdestructuring break as in this one with "Line 55:1 Invalid left-hand side in for-in or for-of" I would expect that ... most of those are the same issue so I am not listing all of them.tryin this test with "Line 36:10 Unexpected token [" which seems more like a case of destructuring not being support in the catch parameter.Disabling
transform-es2015-parametersbreaks a lot of tests and I didn't want to go into it as I was pretty sure I did something wrong on my side, but it seems to not be me 😢 😭 (edit: opened #309). The same goes fortransform-es2015-spreadwhich also makes the tests take forever, probably because of all the failures. On the other hand ... it really shouldn't so I will keep digging in those two babel plugins to see what is going on and report back when I figure what is going on.The text was updated successfully, but these errors were encountered: