Emit source locations manually for more precise tracking. Don't infer
them from emitted bytecode opcodes because that leads to inaccurate
and sometimes surprising results.
Speeds up code generation (although infinitesimally) as a bonus.
Fixes: https://github.com/quickjs-ng/quickjs/issues/236
Introduced in commit 61c8fe6 from last month that moved the callback
into the job queue:
1. It leaked `fre->held_val` when no job was enqueued
2. It fumbled the reference count when enqueuing; JS_EnqueueJob already
takes care of incrementing and decrementing it
Reverts commit 0a70623 from earlier today because that didn't turn out
to be a complete fix.
Fixes: https://github.com/quickjs-ng/quickjs/issues/648
This commit implements resizable ArrayBuffers - RABs for short - and
extends typed arrays (TAs) to support fixed-length and length-tracking
modes.
SharedArrayBuffers (SABs) also support the maxByteLength option now but
I cheated and allocate all memory upfront because atomically resizing
memory allocations is hard and this commit is already big and complex.
The lion's share is updating all the TA prototype methods to deal with
RABs resizing underneath them. Method arguments can be arbitrary objects
with arbitrary .valueOf methods and arbitrary side effects, like...
resizing the RAB we're currently operating on.
Fixes: https://github.com/quickjs-ng/quickjs/issues/477
Deserialized bytecode does not have IC state, i.e., `bc->ic == NULL`.
That may or may not be bug (IMO, it is and we should rebuild the
IC state during deserialization) but, either way, don't segfault.
DRY add_ic_slot() and its call sites in a hopefully NFC manner.
Don't raise a "invalid tag 12" exception when encountering bytecode
and JS_READ_OBJ_BYTECODE is not set, because no one knows what "tag 12"
means without looking it up, not even quickjs maintainers.
Otherwise it's too easy to tie up too many resources (cpu, memory) by
crafting inputs with a very large atom count (up to 4 billion.)
This may need some finetuning. If the limit proves too restrictive for
very large snapshots, we can make it relative to the size of the input.
Check inside the deserializer that const atoms are indeed const, don't
trust the input. The serializer only writes type 0 records for const
atoms but the byte stream may have been corrupted or manipulated.
Overlooked during review of c25aad7 ("Add ability to (de)serialize
symbols")
Found with libfuzzer and it found it _really_ fast. Great tool.
Before this commit it segfaulted, now it throws a SyntaxError.
That's still not correct behavior but better than segfaulting.
To be continued.
Includes a small run-test262 fix to handle Windows line endings.
Refs: https://github.com/quickjs-ng/quickjs/issues/567
Since we don't keep timers sorted by deadline but by insertion order,
the test is flaky in slow environments (GHA seemingly). Increase the
timeouts to give it a bigger chance of success.
ASan / UBSan builds are notoriously slow, so skip the test in those.
The spec says HostMakeJobCallback has to be used on the callback: https://tc39.es/ecma262/multipage/managing-memory.html#sec-finalization-registry-cleanup-callback
That makes the following (arguably contrived) example run forever until
memory is exhausted.
```js
let count = 0;
function main() {
console.log(`main! ${++count}`);
const registry = new FinalizationRegistry(() => {
globalThis.foo = main();
});
registry.register([]);
registry.register([]);
return registry;
}
main();
console.log(count);
```
That is unlike V8, which runs 0 times. This can be explained by the
difference in GC implementations and since FinRec makes GC observable,
here we are!
Fixes: https://github.com/quickjs-ng/quickjs/issues/432
I find it convenient to have access to JS_ReadObject and JS_WriteObject
from JS land. That's precisely the functionality that tests/bjson.c
provides, ergo, move it into quickjs-libc.c
integer conversions:
- improve `u32toa_radix` and `u64toa_radix`, add `i32toa_radix`
- use `i32toa_radix` for small ints in `js_number_toString`
floating point conversions (`js_dtoa`):
- complete rewrite with fewer calls to `snprintf`
- remove `JS_DTOA_FORMAT`, define 4 possible modes for `js_dtoa`
- remove the radix argument in `js_dtoa`
- merge `js_dtoa1` into `js_dtoa`
- add `js_dtoa_infinite` for non finite values
- simplify sign handling
- handle locale specific decimal point transparently
helper function `js_fcvt`:
- simplify `js_fcvt`, remove `js_fcvt1`, reduce overhead
- round up manually instead of using `fesetround(FE_UPWARD)`.
helper function `js_ecvt`:
- document `js_ecvt` and `js_ecvt1` behavior
- avoid redundant `js_ecvt1` calls in `js_ecvt`
- fixed buffer contents, no buffer copies
- simplify decimal point handling
- round up manually instead of using `fesetround(FE_UPWARD)`.
miscellaneous:
- remove `CONFIG_PRINTF_RNDN`. This fixes some of the conversion errors
on Windows. Updated the tests accordingly
- this fixes a v8.sh bug on macOS: `0.5.toFixed(0)` used to produce `0` instead of `1`
- add regression tests, update test_conv unit tests
- add benchmarks for `toFixed`, `toPrecision` and `toExponential` number methods
- benchmarks show all conversions are now 40 to 45% faster (M2)
- use single test in `js_strtod` loop.
- use more explicit `ATOD_xxx` flags
- remove `ATOD_TYPE_MASK`, use `ATOD_WANT_BIG_INT` instead
- remove unused arguments `flags` and `pexponent` in `js_string_to_bigint`
- merge `js_atof` and `js_atof2`, remove `slimb_t *pexponent` argument
- simplify and document `js_atof` parser, remove cumbersome labels,
- simplify `js_parseInt` test for zero radix for `ATOD_ACCEPT_HEX_PREFIX`
- simplify `next_token` number parsing, handle legacy octal in parser only
- simplify `JS_StringToBigInt`, use flags only.
- remove unused `slimb_t exponent` token field
- add number syntax tests
- output more informative error messages in `js_parse_expect`.
The previous code was bogus:
```
return js_parse_error(s, "expecting '%c'", tok);
```
this was causing a bug on `eval("do;")` where `tok` is `TOK_WHILE` (-70, 0xBA)
creating an invalid UTF-8 encoding (lone trailing byte).
This would ultimately have caused a failure in `JS_ThrowError2` if `JS_NewString`
failed when converting the error message to a string if the conversion detected the invalid
UTF-8 encoding and throwed an error (it currently does not, but should).
- test for `JS_NewString` failure in `JS_ThrowError2`
- test for `JS_FreeCString` failure in run-test262.c
- add more test cases
- fix radix conversion rounding code: incrementing the digit
does not work for '9'. We can assume ASCII so it works for
all other digits, especially all letters
- also avoid recomputing the string length
* Add utility functions, improve integer conversion functions
- move `is_be()` to cutils.h
- add `is_upper_ascii()` and `to_upper_ascii()`
- add extensive benchmark for integer conversion variants in **tests/test_conv.c**
- add `u32toa()`, `i32toa()`, `u64toa()`, `i64toa()` based on register shift variant
- add `u32toa_radix()`, `u64toa_radix()`, `i64toa_radix()` based on length_loop variant
- use direct converters instead of `snprintf()`
- copy NaN and Infinity directly in `js_dtoa1()`
- optimize `js_number_toString()` for small integers
- use `JS_NewStringLen()` instead of `JS_NewString()` when possible
- add more precise conversion tests in microbench.js
- disable some benchmark tests for gcc (they cause ASAN failures)
* Fix member accesses for non-decimal numeric literals
e.g. 0x0.a should return undefined, not SyntaxError.
* Remove ineffective non-decimal float parsing code and redundant checks on `is_float && radix != 10`
(The code already wasn't doing anything because of the `is_float` check.)
