pull request commentdenoland/deno
perf: eager poll async ops in Isolate
Very nice work @bartlomieju! This PR has proven significant perf gains in the http server benchmark and throughput benchmark:
However it has increased tail latency. @bartlomieju's subsequent #3128 has mediated it to reasonable levels:
It's worth noting that our benchmark page is generated inside Github Actions, which are perhaps not the best servers to undertake CPU heavy measurements. On the benchmarks page, deno_tcp is shown as performing better than node_tcp. On my own machine, when I benchmark them, node_tcp is still significantly better:
deno_tcp:
> wrk http://127.0.0.1:4500
Running 10s test @ http://127.0.0.1:4500
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 0.89ms 6.85ms 114.82ms 98.91%
Req/Sec 19.33k 2.81k 22.69k 87.00%
384687 requests in 10.00s, 18.71MB read
Requests/sec: 38461.11
Transfer/sec: 1.87MB
node_tcp:
> wrk http://127.0.0.1:4544
Running 10s test @ http://127.0.0.1:4544
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 219.12us 343.96us 10.19ms 97.11%
Req/Sec 24.81k 4.84k 38.29k 80.00%
493490 requests in 10.00s, 24.00MB read
Requests/sec: 49333.27
Transfer/sec: 2.40MB
So there is still work to do. We will be able to upgrade to Tokio's new scheduler in the coming weeks, which should shed some light on how much of this overhead is due to our dependencies.
comment created time in 3 hours
pull request commentdenoland/deno
std: Move fs/path to the top-level
This patch resulted in an increase in the bundle size of file_server.ts ... I'm not sure why.
comment created time in 3 hours
push eventdenoland/deno
commit sha f51dcc12d7a75a677529d63eb53d7a577d5b9289
std: Move fs/path to the top-level (#3100)
push time in 4 hours
PR merged denoland/deno
Moved from https://github.com/denoland/deno_std/pull/624. Closes https://github.com/denoland/deno_std/issues/623.
- Move
fs/pathto the top-level. - Remove
fs/path.ts. - Extract
fs/expand_glob.tsfrompath/glob.ts.
pr closed time in 4 hours
issue closeddenoland/deno_std
To access everything in fs, you have to import from:
fs/mod.tsfs/path.ts(fs/path/mod.ts+fs/path/interface.ts)fs/path/constants.ts
These are all used from outside of fs.
fs/path/mod.ts should re-export everything public in fs/path/constants.ts.
fs/mod.ts should also have to re-export fs/path.ts, but I favour making fs/path top-level.
That would leave:
fs/mod.tspath/mod.ts
Some exports should be moved from path to fs, namely EOL and isWindows... if dependency cycles are okay.
closed time in 4 hours
nayeemrmn
Pull request review commentdenoland/deno
Implemented new open mode in files
export type OpenMode = * @internal */ export function create(filename: string): Promise<File> {- return open(filename, "w+");+ return open(filename, {+ read: true,+ create: true,+ write: true,+ truncate: true+ });+}++/** Check if OpenMode is set to something that is valid.+ * @returns Tuple representing if openMode is valid and error message if it's not+ * @internal+ */+function checkOpenMode(mode: OpenMode): [boolean, string] {+ const allOptionsAreFalse =+ Object.values(mode).filter(val => val == true).length === 0;+ const truncateOptionWithoutWriteAccess = mode.truncate && !mode.write;+ const createOrCreateNewWithoutWriteOrAppend =+ (mode.create || mode.createNew) && !(mode.write || mode.append);+ if (allOptionsAreFalse)+ return [false, "OpenMode require at least one option to be true"];+ if (truncateOptionWithoutWriteAccess)+ return [false, "Truncate option require write accesss"];+ if (createOrCreateNewWithoutWriteOrAppend)+ return [false, "Create or create new requires write or append access"];+ return [true, ""];+}++/** Converts OpenModeLegacy to Legacy+ * @internal+ */+function convertOpenModeLegacy(mode: OpenModeLegacy): OpenMode {+ const r = {+ read: true+ };+ const rplus = {+ read: true,+ write: true+ };+ const w = {+ write: true,+ truncate: true,+ create: true+ };+ const wplus = {+ write: true,+ truncate: true,+ create: true,+ read: true+ };+ const a = {+ append: true,+ create: true+ };+ const aplus = {+ append: true,+ create: true,+ read: true+ };+ const x = {+ write: true,+ createNew: true+ };+ const xplus = {+ write: true,+ createNew: true,+ read: true+ };++ const map = {};+ map["r"] = r;+ map["r+"] = rplus;+ map["w"] = w;+ map["w+"] = wplus;+ map["a"] = a;+ map["a+"] = aplus;+ map["x"] = x;+ map["x+"] = xplus;
I think this can all be shortened up and made more readable by not declaring variables, but rather
const legacyMap = {
"r": { read: true},
"w": { write: true},
// ...
};
function convertOpenModeLegacy(mode: OpenModeLegacy): OpenMode {
return legacyMap[mode];
}
comment created time in 5 hours
Pull request review commentdenoland/deno
Implemented new open mode in files
export type OpenMode = * @internal */ export function create(filename: string): Promise<File> {- return open(filename, "w+");+ return open(filename, {+ read: true,+ create: true,+ write: true,+ truncate: true+ });+}++/** Check if OpenMode is set to something that is valid.+ * @returns Tuple representing if openMode is valid and error message if it's not+ * @internal+ */+function checkOpenMode(mode: OpenMode): [boolean, string] {+ const allOptionsAreFalse =+ Object.values(mode).filter(val => val == true).length === 0;+ const truncateOptionWithoutWriteAccess = mode.truncate && !mode.write;+ const createOrCreateNewWithoutWriteOrAppend =+ (mode.create || mode.createNew) && !(mode.write || mode.append);+ if (allOptionsAreFalse)+ return [false, "OpenMode require at least one option to be true"];+ if (truncateOptionWithoutWriteAccess)+ return [false, "Truncate option require write accesss"];+ if (createOrCreateNewWithoutWriteOrAppend)+ return [false, "Create or create new requires write or append access"];+ return [true, ""];+}++/** Converts OpenModeLegacy to Legacy+ * @internal+ */+function convertOpenModeLegacy(mode: OpenModeLegacy): OpenMode {
I think we should keep it because the "w+" style arguments are very familiar to some. But maybe the name "legacy" isn't so good. OpenModePosix ? OpenModeAbbreviated ?
comment created time in 5 hours
push eventdenoland/deno
commit sha c1b302d76968bb55df24ef360a9fbf71857ae35c
fix: remote jsx/tsx files were compiled as js/ts (#3125)
push time in 5 hours
PR merged denoland/deno
- At resolving remote files, extension are ignored and
.jsx|.tsxfiles were compiled as.jsand.ts, resulting throw compilation error. - To resolve correctly, remote files should be checked by content-type and extension together.
pr closed time in 5 hours
push eventmhvsa/deno
commit sha 4221b90c3febbe03a4b47e47248263741a0fdd4a
perf: eager poll async ops in Isolate (#3046)
commit sha 8114ea525c4bb057f050ebc322a4d2a6ce15dfbc
Upgrade V8 to 7.9.304
commit sha ddbfe9b5aa13e469eb8f2e074352bcdbb9e8c71b
Add debug build to github actions This disabled the tty_tests which seem to be very flaky with the debug build.
commit sha 54db12c4d979bff8f8ebd0dbfe981052c12cb315
perf: fairer ops scheduling (#3128)
commit sha b1685ce0101f9ca16b0eeb6db7ff16ac81301545
Use max-width over max-device-width (#3132) Using max-width because this applies the mobile styles to narrow windows on larger devices as well as full with windows on narrow devices.
commit sha 360714cc6fbe4e1b5e3dbbcaa53d4a0818d76b5b
Merge branch 'master' into partial_reload
commit sha 83ccb746b51532064574c46c06dd467c3e3de4a7
cleanup
push time in 6 hours
Pull request review commentdenoland/deno
[WIP] refactor: error handling in Isolate ops API
impl OpRegistry { op_id: OpId, control: &[u8], zero_copy_buf: Option<PinnedBuf>,- ) -> CoreOp {+ ) -> Result<CoreOp, String> { // Op with id 0 has special meaning - it's a special op that is always // provided to retrieve op id map. The map consists of name to `OpId` // mappings. if op_id == 0 {- return Op::Sync(self.json_map());+ return Ok(Op::Sync(self.json_map())); } - let d = &*self.dispatchers.get(op_id as usize).expect("Op not found!");- d(control, zero_copy_buf)+ let d = match self.dispatchers.get(op_id as usize) {+ Some(handler) => &*handler,+ None => return Err(format!("Unknown op id: {}", op_id)),+ };
Also it would be helpful if you add triple slash doc to this function describing when it fails (i.e. it only fails when op id is wrong).
comment created time in 7 hours
Pull request review commentdenoland/deno
[WIP] refactor: error handling in Isolate ops API
impl OpRegistry { op_id: OpId, control: &[u8], zero_copy_buf: Option<PinnedBuf>,- ) -> CoreOp {+ ) -> Result<CoreOp, String> { // Op with id 0 has special meaning - it's a special op that is always // provided to retrieve op id map. The map consists of name to `OpId` // mappings. if op_id == 0 {- return Op::Sync(self.json_map());+ return Ok(Op::Sync(self.json_map())); } - let d = &*self.dispatchers.get(op_id as usize).expect("Op not found!");- d(control, zero_copy_buf)+ let d = match self.dispatchers.get(op_id as usize) {+ Some(handler) => &*handler,+ None => return Err(format!("Unknown op id: {}", op_id)),+ };
It's a little odd for this low-level function to construct an error string like this. I think it might be better if this function returns Result<CoreOp, ()> or Option<CoreOp> and the caller creates the error message.
comment created time in 7 hours
Pull request review commentdenoland/deno
feat: type check .js file if @ts-check pragma present
class SourceFile { } } +const TS_CHECK_PRAGMA_RE = /^\/\/\s?@ts-check/;++function hasTsCheckPragma(sourceCode: string): boolean {+ return TS_CHECK_PRAGMA_RE.test(sourceCode);+}
If we're in the TS compiler, aren't we already type checking? I feel like this should be a Rust-only feature...?
comment created time in 7 hours
Pull request review commentdenoland/deno
feat: type check .js file if @ts-check pragma present
use url::Url; lazy_static! { static ref CHECK_JS_RE: Regex = Regex::new(r#""checkJs"\s*?:\s*?true"#).unwrap();+ static ref TS_CHECK_PRAGMA_RE: Regex =+ Regex::new(r#"^//\s?@ts-check"#).unwrap();
Isn't this more appropriate in cli/compilers/js.rs ?
comment created time in 7 hours
Pull request review commentdenoland/deno
std: Move fs/path to the top-level
+const { build } = Deno;++export const isWindows = build.os === "win";+export const EOL = isWindows ? "\r\n" : "\n";
@nayeemrmn Maybe you can reduce this patch to just the path renames...
comment created time in 7 hours
Pull request review commentdenoland/deno
std: Move fs/path to the top-level
// Copyright 2018-2019 the Deno authors. All rights reserved. MIT license. -/** EndOfLine character enum */-export enum EOL {- LF = "\n",- CRLF = "\r\n"-}- const regDetect = /(?:\r?\n)/g; /** * Detect the EOL character for string input. * returns null if no newline */-export function detect(content: string): EOL | null {+export function detect(content: string): "\r\n" | "\n" | null {
This seems not so desirable...
comment created time in 7 hours
Pull request review commentdenoland/deno
To get help on the another subcommands (run in this case): ).arg( Arg::with_name("reload") .short("r")+ .min_values(0)+ .takes_value(true)+ .use_delimiter(true)+ .require_equals(true) .long("reload")- .help("Reload source code cache (recompile TypeScript)")+ .help("Reload source code cache (recompile TypeScript). Supports blacklist")+ .value_name("blacklist")+ .long_help("Reload source code cache (recompile TypeScript). Supports blacklist+ --reload+ Reload everything+ --reload=https://deno.land/std+ Reload everything from the standard module+ --reload=https://deno.land/std/fs/utils.ts,https://deno.land/std/fmt/colors.ts+ Reloads only fs/utils and fmt/colors modules")
Ah right. I think "blacklist" makes sense elsewhere in the code, but not here in the user facing docs.
comment created time in 7 hours
PR merged denoland/deno
<!-- Before submitting a PR read https://deno.land/manual.html#contributing -->
pr closed time in 7 hours
push eventdenoland/deno
commit sha 5983507255c8f725ee36ff53f6f8ecb2cb41e607
std/flags: Add test for multiple boolean flags (#3103)
push time in 7 hours
PR opened denoland/deno
merge on approval
pr created time in 8 hours
pull request commentdenoland/deno
Nice patch - thanks! Just a few comments...
comment created time in 8 hours
Pull request review commentdenoland/deno
fn filter_shebang(bytes: Vec<u8>) -> Vec<u8> { } } +fn check_cache_blacklist(url: &Url, black_list: &[String]) -> bool {
Should the black_list argument be a Vec<Url> instead?
comment created time in 8 hours
Pull request review commentdenoland/deno
mod tests { }; } + #[test]+ fn test_cache_blacklist() {+ let args = crate::flags::resolve_urls(vec![+ String::from("http://deno.land/std"),+ String::from("http://github.com/example/mod.ts"),+ String::from("http://fragment.com/mod.ts#fragment"),+ String::from("http://query.com/mod.ts?foo=bar"),+ String::from("http://queryandfragment.com/mod.ts?foo=bar#fragment"),+ ]);++ let url1: Url = "http://deno.land/std/fs/mod.ts".parse().unwrap();+ let url2: Url = "http://github.com/example/file.ts".parse().unwrap();+ let url3: Url = "http://github.com/example/mod.ts".parse().unwrap();+ let url4: Url = "http://github.com/example/mod.ts?foo=bar".parse().unwrap();+ let url5: Url =+ "http://github.com/example/mod.ts#fragment".parse().unwrap();+ let url6: Url = "http://fragment.com/mod.ts".parse().unwrap();+ let url7: Url = "http://query.com/mod.ts".parse().unwrap();+ let url8: Url = "http://fragment.com/mod.ts#fragment".parse().unwrap();+ let url9: Url = "http://query.com/mod.ts?foo=bar".parse().unwrap();+ let url10: Url = "http://queryandfragment.com/mod.ts".parse().unwrap();+ let url11: Url = "http://queryandfragment.com/mod.ts?foo=bar"+ .parse()+ .unwrap();+ let url12: Url = "http://queryandfragment.com/mod.ts#fragment"+ .parse()+ .unwrap();+ let url13: Url =+ "http://query.com/mod.ts?foo=bar#fragment".parse().unwrap();+ let url14: Url = "http://fragment.com/mod.ts?foo=bar#fragment"+ .parse()+ .unwrap();++ let result1 = check_cache_blacklist(&url1, &args);+ let result2 = check_cache_blacklist(&url2, &args);+ let result3 = check_cache_blacklist(&url3, &args);+ let result4 = check_cache_blacklist(&url4, &args);+ let result5 = check_cache_blacklist(&url5, &args);+ let result6 = check_cache_blacklist(&url6, &args);+ let result7 = check_cache_blacklist(&url7, &args);+ let result8 = check_cache_blacklist(&url8, &args);+ let result9 = check_cache_blacklist(&url9, &args);+ let result10 = check_cache_blacklist(&url10, &args);+ let result11 = check_cache_blacklist(&url11, &args);+ let result12 = check_cache_blacklist(&url12, &args);+ let result13 = check_cache_blacklist(&url13, &args);+ let result14 = check_cache_blacklist(&url14, &args);++ assert_eq!(result1, true);+ assert_eq!(result2, false);+ assert_eq!(result3, true);+ assert_eq!(result4, true);+ assert_eq!(result5, true);+ assert_eq!(result6, true);+ assert_eq!(result7, false);+ assert_eq!(result8, true);+ assert_eq!(result9, true);+ assert_eq!(result10, false);+ assert_eq!(result11, true);+ assert_eq!(result12, false);+ assert_eq!(result13, true);+ assert_eq!(result14, true);
Indeed, I would argue defining all these url1, url2, ... variables are not helpful either. It's less code and easier to read if it's written
assert_eq!(true, check_cache_blacklist("http://deno.land/std/fs/mod.ts".parse().unwrap(), &args));
comment created time in 8 hours
Pull request review commentdenoland/deno
mod tests { }; } + #[test]+ fn test_cache_blacklist() {+ let args = crate::flags::resolve_urls(vec![+ String::from("http://deno.land/std"),+ String::from("http://github.com/example/mod.ts"),+ String::from("http://fragment.com/mod.ts#fragment"),+ String::from("http://query.com/mod.ts?foo=bar"),+ String::from("http://queryandfragment.com/mod.ts?foo=bar#fragment"),+ ]);++ let url1: Url = "http://deno.land/std/fs/mod.ts".parse().unwrap();+ let url2: Url = "http://github.com/example/file.ts".parse().unwrap();+ let url3: Url = "http://github.com/example/mod.ts".parse().unwrap();+ let url4: Url = "http://github.com/example/mod.ts?foo=bar".parse().unwrap();+ let url5: Url =+ "http://github.com/example/mod.ts#fragment".parse().unwrap();+ let url6: Url = "http://fragment.com/mod.ts".parse().unwrap();+ let url7: Url = "http://query.com/mod.ts".parse().unwrap();+ let url8: Url = "http://fragment.com/mod.ts#fragment".parse().unwrap();+ let url9: Url = "http://query.com/mod.ts?foo=bar".parse().unwrap();+ let url10: Url = "http://queryandfragment.com/mod.ts".parse().unwrap();+ let url11: Url = "http://queryandfragment.com/mod.ts?foo=bar"+ .parse()+ .unwrap();+ let url12: Url = "http://queryandfragment.com/mod.ts#fragment"+ .parse()+ .unwrap();+ let url13: Url =+ "http://query.com/mod.ts?foo=bar#fragment".parse().unwrap();+ let url14: Url = "http://fragment.com/mod.ts?foo=bar#fragment"+ .parse()+ .unwrap();++ let result1 = check_cache_blacklist(&url1, &args);+ let result2 = check_cache_blacklist(&url2, &args);+ let result3 = check_cache_blacklist(&url3, &args);+ let result4 = check_cache_blacklist(&url4, &args);+ let result5 = check_cache_blacklist(&url5, &args);+ let result6 = check_cache_blacklist(&url6, &args);+ let result7 = check_cache_blacklist(&url7, &args);+ let result8 = check_cache_blacklist(&url8, &args);+ let result9 = check_cache_blacklist(&url9, &args);+ let result10 = check_cache_blacklist(&url10, &args);+ let result11 = check_cache_blacklist(&url11, &args);+ let result12 = check_cache_blacklist(&url12, &args);+ let result13 = check_cache_blacklist(&url13, &args);+ let result14 = check_cache_blacklist(&url14, &args);++ assert_eq!(result1, true);+ assert_eq!(result2, false);+ assert_eq!(result3, true);+ assert_eq!(result4, true);+ assert_eq!(result5, true);+ assert_eq!(result6, true);+ assert_eq!(result7, false);+ assert_eq!(result8, true);+ assert_eq!(result9, true);+ assert_eq!(result10, false);+ assert_eq!(result11, true);+ assert_eq!(result12, false);+ assert_eq!(result13, true);+ assert_eq!(result14, true);
Nit: there's no point in defining all these variables, just do something like:
assert_eq!(true, check_cache_blacklist(&url1, &args));
comment created time in 8 hours
Pull request review commentdenoland/deno
To get help on the another subcommands (run in this case): ).arg( Arg::with_name("reload") .short("r")+ .min_values(0)+ .takes_value(true)+ .use_delimiter(true)+ .require_equals(true) .long("reload")- .help("Reload source code cache (recompile TypeScript)")+ .help("Reload source code cache (recompile TypeScript). Supports blacklist")+ .value_name("blacklist")+ .long_help("Reload source code cache (recompile TypeScript). Supports blacklist+ --reload+ Reload everything+ --reload=https://deno.land/std+ Reload everything from the standard module+ --reload=https://deno.land/std/fs/utils.ts,https://deno.land/std/fmt/colors.ts+ Reloads only fs/utils and fmt/colors modules")
Shouldn't this be a "whitelist" rather than a "blacklist" ?
Also I don't think it's useful to say "Supports blacklist" or "Supports whitelist" - the examples you give are sufficient.
comment created time in 8 hours
push eventdenoland/registry
commit sha ce6493253c11bab60e8c04d7e2297f4f06380864
Add CatYou (#172)
push time in 8 hours
PR merged denoland/registry
I am requesting to add my module CatYou-Deno to the database.
pr closed time in 8 hours
push eventdenoland/registry
commit sha 6c95005ab13e21e89b6c3b9278481cc0efd8b001
add bigfloat (#175)
push time in 8 hours
Pull request review commentdenoland/deno
std: Move fs/path to the top-level
+const { build } = Deno;++export const isWindows = build.os === "win";+export const EOL = isWindows ? "\r\n" : "\n";
Sorry - I’m just hesitant if because this PR is doing a couple of different things... I guess it looks fine - but I would prefer a patch that moves files around to only do that...
comment created time in 8 hours
issue commentdenoland/deno
Good write up. There are a couple bits of machinery implicit in this that are worth mention explicitly:
- There should be a way to register new resources at runtime, obtaining a new rid.
- Resource objects should be up-casted inside the op implementations.
- resource.close() removes a resource from the table.
comment created time in 8 hours
push eventRoryMalcolm/deno
commit sha 94667a9a597d98b1f1b6211cf019272b7824e2da
fix
push time in 19 hours
pull request commentdenoland/deno
There's still one failure:
FAIL customInspectReturnsCorrectHeadersFormat
AssertionError:
[Diff] Left / Right
- "Headers { content-type: application/json, content-length: 1337 }"
+ "Headers { content-type: Application/json, content-length: 1337 }"
comment created time in 19 hours
push eventRoryMalcolm/deno
commit sha af0314f338a1cdb37d446cad4669234d25ce2f5e
fmt
push time in 19 hours
pull request commentdenoland/deno
Does tools/format.py not fix it? It calls prettier - potentially with different configuration than the default prettier.
comment created time in a day
push eventdenoland/deno
commit sha b1685ce0101f9ca16b0eeb6db7ff16ac81301545
Use max-width over max-device-width (#3132) Using max-width because this applies the mobile styles to narrow windows on larger devices as well as full with windows on narrow devices.
push time in a day
PR merged denoland/deno
Using max-width because this applies the mobile styles to narrow windows on larger devices as well as full with windows on narrow devices.
Before
After
pr closed time in a day
issue openeddenoland/deno
https://github.com/denoland/deno/commit/ddbfe9b5aa13e469eb8f2e074352bcdbb9e8c71b
created time in a day
pull request commentdenoland/deno
fix: [jsx] remote jsx/tsx files were compiled as js/ts
@keroxp I've merged the v8 upgrade which should alleviate the failure you were experiencing. I think there is still an issue in this patch though...
comment created time in a day
push eventkeroxp/deno
commit sha 4221b90c3febbe03a4b47e47248263741a0fdd4a
perf: eager poll async ops in Isolate (#3046)
commit sha 8114ea525c4bb057f050ebc322a4d2a6ce15dfbc
Upgrade V8 to 7.9.304
commit sha ddbfe9b5aa13e469eb8f2e074352bcdbb9e8c71b
Add debug build to github actions This disabled the tty_tests which seem to be very flaky with the debug build.
commit sha 0c1bf5c0f82c1bb2d99805ebba0794703c157ac2
Merge branch 'master' into resolve-remote-jsx
push time in a day
push eventdenoland/deno_third_party
commit sha 908bc5a63c3520171d4b5caf6ec1f9c66633af70
Upgrade V8 to 7.9.304
push time in a day
push eventdenoland/deno
commit sha 8114ea525c4bb057f050ebc322a4d2a6ce15dfbc
Upgrade V8 to 7.9.304
commit sha ddbfe9b5aa13e469eb8f2e074352bcdbb9e8c71b
Add debug build to github actions This disabled the tty_tests which seem to be very flaky with the debug build.
push time in a day
issue closeddenoland/deno
deno fmt fails with error of Illegal instruction
🙁
tmnk@buildDeno:~/deno$ ./target/debug/deno fmt badly_formatted.js
Download https://deno.land/std@8c90bd9/prettier/vendor/standalone.js
Download https://deno.land/std@8c90bd9/prettier/vendor/parser_typescript.js
Download https://deno.land/std@8c90bd9/prettier/vendor/parser_markdown.js
#
# Fatal error in ../../core/libdeno/v8/src/parsing/preparser.h, line 1076
# Debug check failed: !parsing_module_.
#
#
#
#FailureMessage Object: 0x7eff75515cc0
==== C stack trace ===============================
./target/debug/deno(+0x286acb3) [0x55aadeed1cb3]
./target/debug/deno(+0x273e52b) [0x55aadeda552b]
./target/debug/deno(+0x1e6e648) [0x55aade4d5648]
./target/debug/deno(+0x1e6e375) [0x55aade4d5375]
./target/debug/deno(+0x1adf30f) [0x55aade14630f]
./target/debug/deno(+0x1a7927e) [0x55aade0e027e]
./target/debug/deno(+0x1a7b081) [0x55aade0e2081]
./target/debug/deno(+0x1a7abd0) [0x55aade0e1bd0]
./target/debug/deno(+0x1ad1346) [0x55aade138346]
./target/debug/deno(+0x1acf856) [0x55aade136856]
./target/debug/deno(+0x1ad6f14) [0x55aade13df14]
./target/debug/deno(+0x1ad68a6) [0x55aade13d8a6]
./target/debug/deno(+0x1ad4ee4) [0x55aade13bee4]
./target/debug/deno(+0x1ad63f0) [0x55aade13d3f0]
./target/debug/deno(+0x1ad4fea) [0x55aade13bfea]
./target/debug/deno(+0x1ad2753) [0x55aade139753]
./target/debug/deno(+0x1adf556) [0x55aade146556]
./target/debug/deno(+0x1addc1e) [0x55aade144c1e]
./target/debug/deno(+0x1a7927e) [0x55aade0e027e]
./target/debug/deno(+0x1a7b081) [0x55aade0e2081]
./target/debug/deno(+0x1a7abd0) [0x55aade0e1bd0]
./target/debug/deno(+0x1ad1346) [0x55aade138346]
./target/debug/deno(+0x1acf856) [0x55aade136856]
./target/debug/deno(+0x1ad6f14) [0x55aade13df14]
./target/debug/deno(+0x1ad68a6) [0x55aade13d8a6]
./target/debug/deno(+0x1ad4ee4) [0x55aade13bee4]
./target/debug/deno(+0x1ad45b4) [0x55aade13b5b4]
./target/debug/deno(+0x1aced95) [0x55aade135d95]
./target/debug/deno(+0x1ad6f78) [0x55aade13df78]
./target/debug/deno(+0x1ad68a6) [0x55aade13d8a6]
./target/debug/deno(+0x1ad4ee4) [0x55aade13bee4]
./target/debug/deno(+0x1ad63f0) [0x55aade13d3f0]
./target/debug/deno(+0x1adb37b) [0x55aade14237b]
./target/debug/deno(+0x1acc833) [0x55aade133833]
./target/debug/deno(+0x1a7927e) [0x55aade0e027e]
./target/debug/deno(+0x1a7b081) [0x55aade0e2081]
./target/debug/deno(+0x1a7abd0) [0x55aade0e1bd0]
./target/debug/deno(+0x1ad1346) [0x55aade138346]
./target/debug/deno(+0x1acf856) [0x55aade136856]
./target/debug/deno(+0x1ad6f14) [0x55aade13df14]
./target/debug/deno(+0x1ad68a6) [0x55aade13d8a6]
./target/debug/deno(+0x1ad4ee4) [0x55aade13bee4]
./target/debug/deno(+0x1ad45b4) [0x55aade13b5b4]
./target/debug/deno(+0x1aced95) [0x55aade135d95]
./target/debug/deno(+0x1ad6f78) [0x55aade13df78]
./target/debug/deno(+0x1ad68a6) [0x55aade13d8a6]
./target/debug/deno(+0x1ad4ee4) [0x55aade13bee4]
./target/debug/deno(+0x1ad63f0) [0x55aade13d3f0]
./target/debug/deno(+0x1adb37b) [0x55aade14237b]
./target/debug/deno(+0x1acc833) [0x55aade133833]
./target/debug/deno(+0x1a7927e) [0x55aade0e027e]
./target/debug/deno(+0x1a7a423) [0x55aade0e1423]
./target/debug/deno(+0x1a798e0) [0x55aade0e08e0]
./target/debug/deno(+0x284c271) [0x55aadeeb3271]
./target/debug/deno(+0x284155b) [0x55aadeea855b]
./target/debug/deno(+0x2859c4d) [0x55aadeec0c4d]
./target/debug/deno(+0x2858284) [0x55aadeebf284]
./target/debug/deno(+0x2856a34) [0x55aadeebda34]
./target/debug/deno(+0x2855e46) [0x55aadeebce46]
./target/debug/deno(+0x28548e0) [0x55aadeebb8e0]
./target/debug/deno(+0x285bf35) [0x55aadeec2f35]
./target/debug/deno(+0x2857615) [0x55aadeebe615]
Illegal instruction (core dumped)
tmnk@buildDeno:~/deno$ cat badly_formatted.js
console.log(
"Hello World"
)
tmnk@buildDeno:~/deno$ cat /proc/version
Linux version 5.0.0-1018-azure (buildd@lgw01-amd64-048) (gcc version 7.4.0 (Ubuntu 7.4.0-1ubuntu1~18.04.1)) #19~18.04.1-Ubuntu SMP Wed Aug 21 05:13:05 UTC 2019
closed time in a day
7k8m
push eventdenoland/deno
commit sha 9beacd108116beb353db55e5fd7d51015241f862
Add debug build to github actions This disabled the tty_tests which seem to be very flaky with the debug build.
push time in a day
push eventdenoland/deno
commit sha 49632bc2f70c7eefd60935426f7792887a79dd2c
bump
push time in 2 days
push eventdenoland/deno
commit sha 68b2eec95aacccf2cca61572d566aaf477c26157
disable fmt_tests
push time in 2 days
push eventdenoland/deno
commit sha 3b706d325482fb4c05c1310fdb4d8d43ffd9d29f
flaky test
push time in 2 days
push eventdenoland/deno
commit sha 79f423009f357a5347df5b4d4b547342c6169584
disable flaky test
push time in 2 days
issue commentdenoland/deno
It might be simpler to just restructure our URL class to not have _parts but rather members? Is that how Node's is implemented?
comment created time in 2 days
push eventdenoland/deno
commit sha 6135e629906f005115a6e7edf38d015470232aac
Add debug build to github actions
push time in 2 days
push eventdenoland/deno
commit sha 4221b90c3febbe03a4b47e47248263741a0fdd4a
perf: eager poll async ops in Isolate (#3046)
push time in 2 days
PR merged denoland/deno
Revamp of #2779 with new dispatcher infra.
I get ridiculous result on my machine...
this PR debug
third_party/wrk/mac/wrk -d 10s --latency http://127.0.0.1:4544
Running 10s test @ http://127.0.0.1:4544
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 52.80ms 185.40ms 1.28s 92.56%
Req/Sec 16.56k 4.16k 26.77k 74.26%
Latency Distribution
50% 283.00us
75% 291.00us
90% 73.22ms
99% 951.90ms
332831 requests in 10.10s, 16.19MB read
Requests/sec: 32954.40
Transfer/sec: 1.60MB
master debug
third_party/wrk/mac/wrk -d 10s --latency http://127.0.0.1:4544
Running 10s test @ http://127.0.0.1:4544
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 584.31us 191.76us 8.23ms 96.67%
Req/Sec 8.67k 777.07 8.99k 94.06%
Latency Distribution
50% 557.00us
75% 561.00us
90% 579.00us
99% 0.99ms
174285 requests in 10.10s, 8.48MB read
Requests/sec: 17255.87
Transfer/sec: 859.42KB
this PR release
third_party/wrk/mac/wrk -d 10s --latency http://127.0.0.1:4544
Running 10s test @ http://127.0.0.1:4544
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 104.40us 62.98us 2.97ms 86.85%
Req/Sec 47.33k 2.91k 50.13k 86.50%
Latency Distribution
50% 97.00us
75% 123.00us
90% 165.00us
99% 299.00us
941792 requests in 10.00s, 45.81MB read
Requests/sec: 94179.51
Transfer/sec: 4.58MB
master release
third_party/wrk/mac/wrk -d 10s --latency http://127.0.0.1:4544
Running 10s test @ http://127.0.0.1:4544
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 122.79us 32.44us 1.95ms 96.50%
Req/Sec 39.58k 1.49k 40.66k 97.03%
Latency Distribution
50% 119.00us
75% 122.00us
90% 125.00us
99% 264.00us
795753 requests in 10.10s, 38.70MB read
Requests/sec: 78785.96
Transfer/sec: 3.83MB
That's ~90% increase for debug and ~20% for release. You can see that latency got worse, but that's promising results.
CC @ry
pr closed time in 2 days
push eventdenoland/deno
commit sha 6a2734ce3bc01e91fd4cef0ef1b45c19e5fd55a7
Add debug build to github actions
push time in 2 days
push eventdenoland/deno
commit sha 28aafd8a97179cde824a825bf8b346f7fcd5499c
Add debug build to github actions
push time in 2 days
issue commentdenoland/deno
deno fmt fails with error of Illegal instruction
This error first appeared in bed7034fc44a8d20e11a85b6092e7f0f34f43ff8.
comment created time in 2 days
issue commentsteveicarus/iverilog
$fatal should result in a non-zero exit code
It's standard in unix to indicate error with non-zero exit code and given that iverilog also does not print to stderr on $fatal means that automated tests require extra code to parse the output in order to utilize iverilog.
I'm not sure which is the best appropriate specifciation but at least in this one it states "$fatal is a run-time fatal, which shall terminates the simulation with an error code."
comment created time in 2 days
issue commentdenoland/deno
deno fmt fails with error of Illegal instruction
We really need to add debug builds to CI...
comment created time in 2 days
issue closeddenoland/deno
Hi,
As described in this issue, when I run cargo test I receive this output:
FAILED (errors=1)
', cli/tests/util/mod.rs:34:5
stack backtrace:
0: backtrace::backtrace::libunwind::trace
at /Users/vsts/.cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.34/src/backtrace/libunwind.rs:88
1: backtrace::backtrace::trace_unsynchronized
at /Users/vsts/.cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.34/src/backtrace/mod.rs:66
2: std::sys_common::backtrace::_print
at src/libstd/sys_common/backtrace.rs:47
3: std::sys_common::backtrace::print
at src/libstd/sys_common/backtrace.rs:36
4: std::panicking::default_hook::{{closure}}
at src/libstd/panicking.rs:200
5: std::panicking::default_hook
at src/libstd/panicking.rs:211
6: std::panicking::rust_panic_with_hook
at src/libstd/panicking.rs:477
7: std::panicking::continue_panic_fmt
at src/libstd/panicking.rs:384
8: std::thread::local::fast::Key<T>::try_initialize
9: integration_tests::util::run_python_script
at cli/tests/util/mod.rs:34
10: integration_tests::fmt_test
at cli/tests/integration_tests.rs:32
11: integration_tests::fmt_test::{{closure}}
at cli/tests/integration_tests.rs:30
12: core::ops::function::FnOnce::call_once
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libcore/ops/function.rs:235
13: <alloc::boxed::Box<F> as core::ops::function::FnOnce<A>>::call_once
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/liballoc/boxed.rs:787
14: __rust_maybe_catch_panic
at src/libpanic_unwind/lib.rs:80
15: std::panicking::try
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libstd/panicking.rs:275
16: std::panic::catch_unwind
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libstd/panic.rs:394
17: test::run_test::run_test_inner::{{closure}}
at src/libtest/lib.rs:1408
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
---- js_unit_tests stdout ----
thread 'js_unit_tests' panicked at 'called `Result::unwrap()` on an `Err` value: "PoisonError { inner: .. }"', src/libcore/result.rs:1084:5
stack backtrace:
0: backtrace::backtrace::libunwind::trace
at /Users/vsts/.cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.34/src/backtrace/libunwind.rs:88
1: backtrace::backtrace::trace_unsynchronized
at /Users/vsts/.cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.34/src/backtrace/mod.rs:66
2: std::sys_common::backtrace::_print
at src/libstd/sys_common/backtrace.rs:47
3: std::sys_common::backtrace::print
at src/libstd/sys_common/backtrace.rs:36
4: std::panicking::default_hook::{{closure}}
at src/libstd/panicking.rs:200
5: std::panicking::default_hook
at src/libstd/panicking.rs:211
6: std::panicking::rust_panic_with_hook
at src/libstd/panicking.rs:477
7: std::panicking::continue_panic_fmt
at src/libstd/panicking.rs:384
8: rust_begin_unwind
at src/libstd/panicking.rs:311
9: std::panicking::begin_panic
10: std::panicking::begin_panic
11: core::result::Result<T,E>::unwrap
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libcore/result.rs:852
12: deno_cli::test_util::http_server
at cli/test_util.rs:62
13: integration_tests::js_unit_tests
at cli/tests/integration_tests.rs:38
14: integration_tests::js_unit_tests::{{closure}}
at cli/tests/integration_tests.rs:37
15: core::ops::function::FnOnce::call_once
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libcore/ops/function.rs:235
16: <alloc::boxed::Box<F> as core::ops::function::FnOnce<A>>::call_once
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/liballoc/boxed.rs:787
17: __rust_maybe_catch_panic
at src/libpanic_unwind/lib.rs:80
18: std::panicking::try
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libstd/panicking.rs:275
19: std::panic::catch_unwind
at /rustc/625451e376bb2e5283fc4741caa0a3e8a2ca4d54/src/libstd/panic.rs:394
20: test::run_test::run_test_inner::{{closure}}
at src/libtest/lib.rs:1408
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
My cargo --version is cargo 1.38.0 (23ef9a4ef 2019-08-20) & rustc --version is rustc 1.38.0 (625451e37 2019-09-23).
I also get this error earlier in the format test (python2 tools/format.py passes) and that is where the panic occurs:
Traceback (most recent call last):
File "tools/fmt_test.py", line 31, in test_fmt
quiet=True)
File "/Users/usrname/Documents/deno/tools/util.py", line 99, in run_output
sys.exit(retcode)
SystemExit: -4
I don't know enough about the logic of util.py to know why this is being thrown from L99 but it's in the run_output function.
closed time in 2 days
RoryMalcolm
issue commentdenoland/deno
Looks like a duplicate of https://github.com/denoland/deno/issues/3122
We'll move the discussion there. Thanks @RoryMalcolm !
comment created time in 2 days
issue commentdenoland/deno
@bartlomieju I agree.
@RoryMalcolm It would be helpful if you could provide the complete output of the cargo test, not just the error.
comment created time in 2 days
issue commentdenoland/deno
Can you provide the output of tools/format.py ?
comment created time in 2 days
Pull request review commentdenoland/deno
fix: [jsx] remote jsx/tsx files were compiled as js/ts
mod tests { map_content_type(Path::new("foo/bar.ts"), Some("foo/bar")), msg::MediaType::Unknown );+ assert_eq!(+ map_content_type(+ Path::new("foo/bar.tsx"),+ Some("application/typescript")+ ),+ msg::MediaType::TSX+ );+ assert_eq!(+ map_content_type(+ Path::new("foo/bar.tsx"),+ Some("application/javascript")+ ),+ msg::MediaType::TSX+ );+ assert_eq!(+ map_content_type(+ Path::new("foo/bar.tsx"),+ Some("application/x-typescript")+ ),+ msg::MediaType::TSX+ );+ assert_eq!(+ map_content_type(+ Path::new("foo/bar.jsx"),+ Some("application/javascript")+ ),+ msg::MediaType::JSX+ );+ assert_eq!(+ map_content_type(+ Path::new("foo/bar.jsx"),+ Some("application/x-typescript")+ ),+ msg::MediaType::JSX+ );
👍
comment created time in 2 days
issue commentdenoland/deno
There may be a random python process running in the background. Can you run
ps aux | grep python
Is there something related to deno in there? A script called http_server.py ...
It may be complicating the debugging here. If it exists, please kill it and try running RUST_BACKTRACE=1 cargo test >& output.txt
Thanks for doing this : )
comment created time in 2 days
issue commentdenoland/deno
Do you have rustfmt installed? What does which rustfmt return?
comment created time in 2 days
issue commentdenoland/deno
@RoryMalcolm Thanks! What version of python are you using?
comment created time in 2 days
issue commentdenoland/deno_std
console.log(headers) doesn't print useful info
@RoryMalcolm It would be very helpful if you run this:
RUST_BACKTRACE=1 cargo test
and open an issue in https://github.com/denoland/deno with the output.
comment created time in 2 days
issue commentdenoland/deno_std
console.log(headers) doesn't print useful info
@RoryMalcolm I'd be interested to hear what issue you ran into.
comment created time in 2 days
issue openedsteveicarus/iverilog
$fatal should result in a non-zero exit code
module tb;
initial begin
#1 if (1) $fatal(1);
#1 $finish;
end
endmodule
I would expect the executable generated with iverilog to have exit code 1, but it has exit code 0.
I'm using Icarus Verilog version 11.0 (devel) (s20150603-643-gb43fcccc)
created time in 2 days
Pull request review commentdenoland/deno
fix: [jsx] remote jsx/tsx files were compiled as js/ts
fn map_content_type(path: &Path, content_type: Option<&str>) -> msg::MediaType { | "text/typescript" | "video/vnd.dlna.mpeg-tts" | "video/mp2t"- | "application/x-typescript" => msg::MediaType::TypeScript,+ | "application/x-typescript" => {+ map_js_like_extension(path, msg::MediaType::TypeScript)+ } "application/javascript" | "text/javascript" | "application/ecmascript" | "text/ecmascript"- | "application/x-javascript" => msg::MediaType::JavaScript,+ | "application/x-javascript" => {+ map_js_like_extension(path, msg::MediaType::JavaScript)+ }
I think there should be a unit test added in this file to test this change...
comment created time in 2 days
Pull request review commentdenoland/deno
fix: [jsx] remote jsx/tsx files were compiled as js/ts
+import "https://raw.githubusercontent.com/denoland/deno/master/cli/tests/047_jsx_test.jsx";Use "http://127.0.0.1/cli/tests/047_jsx_test.jsx" here and elsewhere.
comment created time in 2 days
pull request commentdenoland/deno
Vendor Acorn; Use Acorn parser in REPL
How much larger does this make the deno executable in release mode?
comment created time in 3 days
push eventdenoland/deno
commit sha 605659535794ca0d8fe3ee4ea5857b418d7ce091
fix: [tls] op_dial_tls is not registerd and broken (#3121)
push time in 3 days
PR merged denoland/deno
dialTLSwas broken on 97d8498d- Or broken from v0.20.0? v0.19.0's
dialTLSworks fine.
- Or broken from v0.20.0? v0.19.0's
- added more accurate test.
pr closed time in 3 days
push eventdenoland/deno
commit sha b3331e81d920ac6654545094c31bfe07abc750d5
support --allow-net=:4500 (#3115)
push time in 4 days
PR merged denoland/deno
Closes #2405
This PR adds functionality to expand host passed to --allow-net. If you pass "bare port" then it will be expanded to: 0.0.0.0:port, 127.0.0.1:port and localhost:port.
pr closed time in 4 days
push eventdenoland/deno
commit sha 5f5583cf19349d19f717f65312558588825ad7b8
fix: [prettier] deno fmt should format jsx/tsx files (#3118)
push time in 4 days
PR merged denoland/deno
- Made
deno fmtformat.jsx.tsxfiles.
pr closed time in 4 days
issue closeddenoland/deno
After https://github.com/denoland/deno/pull/2825 lands, I want to change the layout of our source tree. #2825 adds two new crates, bringing our total to 4. I think we need to move them all into a special src directory to make grepping easier, and easier for newbies to navigate.
After #2825, the //js directory is only used by cli_snapshots, I think it should be moved to a subdirectory of that crate.
Thus I propose
//src/
//src/core/
//src/core/libdeno/
//src/cli/
//src/cli/ops
//src/cli_snapshots/
//src/cli_snapshots/js/
//src/deno_typescript/
deno_typescript is a long name - I'd like to use something shorter. But I'm not sure what a good name would be that is descriptive yet short.. I think we can punt on that.
Also I am strongly considering merging the deno_std tree into this repo. I think //src/std would be logical. I will do that in a separate PR, but just want to mention it.
closed time in 4 days
ry
issue commentdenoland/deno
Done. I’m not going to pursue the src dir.
comment created time in 4 days
issue commentdenoland/deno
Thanks for the good outline. I agree with splitting the workers.
I do kinda think it might be better to complete the op crate refactor before splitting up the workers.
comment created time in 4 days
issue commentdenoland/deno
Major features necessary for 1.0
I'm removing fs-events as a blocker for 1.0. It's certainly important but a must-have.
comment created time in 5 days
Pull request review commentdenoland/deno
std: Move fs/path to the top-level
+const { build } = Deno;++export const isWindows = build.os === "win";+export const EOL = isWindows ? "\r\n" : "\n";
I apologize for being so nit-picky about this...
- These 4 lines could be one:
export const EOL = Deno.build.os === "win" ? "\r\n" : "\n"; - Do we need a standalone module for this one constant? What about moving this to
std/fs/eol.ts?
comment created time in 5 days
push eventdenoland/deno
commit sha 97d8498d46bbfb7bb16eeb3a69565d2aeda1e5aa
Add init methods for each op module (#3087)
push time in 5 days
PR merged denoland/deno
Towards #2180
pr closed time in 5 days
PR merged denoland/deno
Fixes #3104.
I still think it should be an entry in DenoDir, though.
pr closed time in 5 days
push eventdenoland/deno
commit sha 04ed8d0531b7c2c308b28b742f5c332345d97065
Ensure DENO_DIR when saving the REPL history (#3106)
push time in 5 days
issue closeddenoland/deno
failure: start REPL without an existing DENO_DIR
➜ try deno
>
Unable to save REPL history: "/home/jason/.cache/deno/deno_history.txt" directory does not exist
closed time in 5 days
CGQAQ
Pull request review commentdenoland/deno
Ensure DENO_DIR when saving the REPL history
def test_lexical_scoped_variable(self): self.assertEqual(code, 0) def test_missing_deno_dir(self):+ deno_dir = "nonexistent" new_env = os.environ.copy()- new_env["DENO_DIR"] = os.path.abspath("doesnt_exist")- out, err, code = self.input("'noop'", exit=False, env=new_env)- self.assertEqual(out, "noop\n")- self.assertTrue(err.startswith("Unable to save REPL history:"))+ new_env["DENO_DIR"] = deno_dir+ out, err, code = self.input("1", exit=False, env=new_env)+ self.assertTrue(os.path.isdir(deno_dir))+ shutil.rmtree(deno_dir)+ self.assertEqual(out, "1\n")+ self.assertEqual(err, "")
👍
comment created time in 5 days
pull request commentdenoland/deno
Ensure DENO_DIR when saving the REPL history
I still think it should be an entry in DenoDir, though.
What do you mean by that?
comment created time in 5 days
pull request commentdenoland/deno
perf: Greedy poll async ops in Isolate
Looks good - I'm in favor of landing this. Nice work! I look forward to seeing the benchmarks page updated.
I want to give @piscisaureus a chance to review since this is very sensitive code.
comment created time in 5 days
Pull request review commentdenoland/deno
perf: Greedy poll async ops in Isolate
void deno_respond(Deno* d_, void* user_data, deno_op_id op_id, deno_buf buf) { if (d->current_args_ != nullptr) { // Synchronous response. // Note op_id is not passed back in the case of synchronous response.- if (buf.data_ptr != nullptr) {+ if (buf.data_ptr != nullptr && buf.data_len > 0) {
I think this alone is a good optimization.
comment created time in 5 days