This slightly degrades the performance of CJS export analysis on
subsequent runs because I changed it to no longer cache in the DENO_DIR
with this PR (denort now properly has no idea about the DENO_DIR). We'll
have to change it to embed this data in the binary and that will also
allow us to get rid of swc in denort (will do that in a follow-up PR).
The TLS start sequence has been broken since #26661 because of the way
how we wrap TCP handle to create TLS handle.
#26661 introduced happy-eyeballs algorithm and some connection could be
dropped because of happy-eyeball attempt timeout. The current
implementation doesn't consider that case and it could start TLS
handshake with timed out TCP connection. That caused #27652 .
This PR fixes it by changing the initialization steps. Now `wrapHandle`
of TLSSocket set up `afterConnectTls` callback in TCP handle, and
`afterConnect` of TCP handle calls it at `connect` event timing if it
exists. This avoids starting TLS session with timed out connection.
closes #27652
Fixes #26224.
Fixes #27042.
There were three bugs here:
- we were only resolving `/// <reference types` directives starting with
`npm:`, which meant we failed to resolve bare specifiers (this broke the
`/// <reference types="vite/client">` directive in most of the vite
templates)
- the `$node_modules` workaround caused us to fail to read files for
tsc. For instance tsc would construct new paths based on specifiers
containing `$node_modules`, and since we hadn't created those we weren't
mapping them back to the original (this broke some type resolution
within `vite/client`)
- our separation of `ImportMeta` across node and deno globals in tsc
meant that npm packages couldn't augment `ImportMeta` (this broke
`vite/client`'s augmentation to add `import.meta.env` and others)
After this, the only remaining issue in the vanilla vite template is our
error on `/vite.svg` (which is an ambient module), and I'll look into
that next.
Fixes https://github.com/denoland/deno/issues/27062
In the LSP we were passing `npm` specifiers to TSC as roots, but TSC
needs fully resolved specifiers (like the actual file path).
In `deno check` we were often excluding the specifiers entirely from the
roots.
In both cases, we need to resolve the specifiers fully and then pass
them to tsc
This was using the lockfile and esm.sh changed breaking the lockfile. We
could pin to a specific esm.sh version, but ideally we shouldn't have
the test suite dependent on remote servers.
This PR changes the underlying buffer backed AST format we use for
JavaScript-based linting plugins. It adds support for various new types,
makes traversal code a lot easier and is more polished compared to
previous iterations.
Here is a quick summary (in no particular order):
- Node prop data is separate from traversal, which makes traversal code
so much easier to reason about. Previously, it was interleaved with node
prop data
- spans are in a separate table as well, as they are rarely needed.
- schema is separate from SWC conversion logic, which makes
- supports recursive plain objects
- supports numbers
- supports bigint
- supports regex
- adds all SWC nodes
Apologies, this is kinda a big PR, but it's worth it imo.
_Marking as draft because I need to update some tests tomorrow._
Fixes #27343
Currently the node:child_process polyfill is always passing the full
parent environment to all spawned subprocesses. In the case where
`options.env` is provided those keys are overridden but the rest of the
parent environment is still passed through.
On Node the behaviour is for child processes to only inherit the parent
environment when `options.env` isn't specified. When `options.env` is
specified the child process inherits only those keys.
This PR updates the internal node child_process polyfill so that the
`clearEnv` argument is set to true when spawning the subprocess to
prevent the parent environment always being inherited by default. It
also fixes an issue where `normalizeSpawnArguments` wasn't returning the
`env` option if `options.env` was unset.
See the comment
https://github.com/denoland/deno/pull/25470#issuecomment-2435077722 for
the reason why we do this workaround to make `make-fetch-happen` work in
Deno
This PR applies the same workaround to `npm-check-updates` package.
`npm-check-updates` internally uses
[`npm-registry-fetch`](https://www.npmjs.com/package/npm-registry-fetch)
which uses
[`make-fetch-happen`](https://www.npmjs.com/package/make-fetch-happen)
(the problematic package) for making http request to npm registry.
The detection of `make-fetch-happen` doesn't work for
`npm-check-updates` because we use call stack at `net.Socket`
constructor to check if it's called from `make-fetch-happen`, but
`npm-check-updates` bundles its dependency and the check doesn't work.
This PR adds the check of `npm-check-updates` string in call stack in
net.Socket constructor to trigger the workaroud.
closes #27629
Fixes https://github.com/denoland/deno/issues/25762. Note that some of
the things in that issue are not resolved (vite/client types not working
properly which has other root causes), but the wildcard module
augmentation specifically is fixed by this.
We were telling TSC that files with unknown media types had an extension
of `.js`, so the ambient module declarations weren't applying. Instead,
just don't resolve them, so the ambient declaration applies.
The selector splitting code that's used for JS linting plugins didn't
properly account for selectors being a single character. This can happen
in the case of `*`.
Instead of comparing against the length, we'll now check if the
remaining string portion is not empty, which is more robust. It also
allows us to detect trailing whitespace, which we didn't before.
A QUIC endpoint is a UDP socket which multiplexes QUIC sessions, which
may be initiated in either direction. This PR exposes endpoints and
moves things around as needed.
Now that endpoints can be reused between client connections, we have a
way to share tls tickets between them and allow 0rtt. This interface
currently works by conditionally returning a promise.
Also cleaned up the rust op names, fixed some lingering problems in the
data transmission, and switched to explicit error types.
The error message is currently `Bad Resource ID`. This commit changes it to
`Cannot read request body as underlying resource unavailable`
closes #27133
Instead of hard erroring, we now surface module not found errors as
TypeScript diagnostics (we have yet to show the source code of the
error, but something we can improve over time).
When running selectors for JS linting plugins we would error when
encountering an unknown attribute name:
```js
// selector
Foo[non-existant]
// error
Error: Missing string id: <number>
```
This was caused by using `0` as the invalid marker, but also overloading
`0` with an actual node type. So the fix is to reserve `0` as the
invalid marker and move the property type to the next index.
Currently we only supports 7 ciphers (`aes-(128|192|256)-ecb` and
`aes-(128|256)-(cbc|gcm)`) in `node:crypto`, but `crypto.getCiphers`
returns other supported cipher names. That confuses `npm:openpgp`
package and causes https://github.com/denoland/deno/issues/26875.
This PR makes `getCiphers` return actually supported cipher names.
With this change, the example given in #26875 can create private and
public key files.
closes #26875
fixes #27132
When the body resource is unavailable when start reading it, the error
message is `Bad Resource ID` and that doesn't tell what's wrong very
well.
This PR changes that error message to `Cannot read body as underlying
resource unavailable`
