This message doesn't require a response of any kind, but handling
it here will prevent an unknown PDU error which kills the session.
Co-authored-by: Benoît Cortier <bcortier@proton.me>
even in cases where the server doesn't send a Capabilities Request
as expected. This mimics the logic in FreeRDP, and empirically appears
to fix a bug where DVC channels behave irregularly.
- Extend ReadCursor with methods to read and try reading i64 values in both little and big endian formats.
- Update RdpdrPdu to handle ClientDriveSetInformationResponse, reflecting the addition of ServerDriveSetInformationRequest handling in the RDPDR (Remote Desktop Protocol: Device Redirection) protocol.
- Implement decoding and encoding of various FileInformationClass types such as FileEndOfFileInformation, FileDispositionInformation, FileRenameInformation, and FileAllocationInformation.
- Refactor the existing code to improve the handling of different FileInformationClassLevels and FileSystemInformationClassLevels.
- Add handling for ServerDriveNotifyChangeDirectoryRequest in ServerDriveIoRequest, supporting the specific case when MinorFunction is set to IRP_MN_NOTIFY_CHANGE_DIRECTORY.
Adds support for native cursors for `IronRDP-web`.
- Uses CSS `cursor` property to set the custom cursor (which we encode as Base64 data URL
- Software rendering is still in place and could be enabled via the config flag
- "Inverted colors" are implemented for native cursors as a "checkerboard pattern" (As in FreeRDP)
- Native cursors have size **limitation** - _Any remote cursor bigger than 32x32 will be scaled down._ This is due to browser limitations described [here](https://chromestatus.com/feature/5825971391299584) and [here](https://bugs.chromium.org/p/chromium/issues/detail?id=880863)
Closes#250
Instead of directly depending on the unmaintained rdp-rs to decompress
RLE tiles dynamically, a few snapshots are saved to test against.
The benefit is to remove a lot of outdated dependencies from our
(developement) dependency graph.
Hiding the lifetimes is making the code less obvious.
That’s not a problem for types that I know about intimately such as
`ReadCursor` or `WriteCursor`, but I actually found myself surprised
more than once when reading code I didn’t authored, discovering later
there was in fact a hidden lifetime parameter.
I expect this problem to be worse for someone not familiar with
our codebase.
I understand this lint is "allow" by default because in some cases
it leads to writing unergonomic ugly code when a type has many generic
lifetimes parameters:
```
TyCtxt<'_, '_, '_>
```
However we _never_ work with more than one generic lifetime parameter in
IronRDP codebase, so it seems to me that the tradeoff towards clarity is
worth it, in our case.
Adds handling for `DR_CORE_SERVER_CLIENTID_CONFIRM` and `DR_CORE_DEVICELIST_ANNOUNCE`/`DR_CORE_DEVICELIST_ANNOUNCE_REQ`.
The next steps in the rdpdr initialization sequence.
Instead of copying and sending image buffers to JavaScript, the WASM
module now draws into the canvas by itself. This removes some overhead
associated with the previous approach and open the door for further
optimizations.
In order to achieve good performance, the newest API of
`softbuffer@0.3.0` is used: the "owned buffer" that can be written into
by us with direct access and `present_with_damage` to apply partial
updates. The presentation itself is currently not yet "no-copy" in
the case of the web backend because the current API of `softbuffer` is
expecting a pixel buffer in the BGRX format while the underlying canvas
can only takes RGBA pixels. There is an open issue for this.
There was a bug with the `present_with_damage` implementation for the
web backend. I fixed the issue and opened a PR to upstream the patch.
The cargo dependency patch will be removed once the fix is published
on crates.io.
Issue: ARC-164
FastPath pointer messages handling:
- Add pointer messages handling in `ironrdp-session`, `ironrdp-client` and `ironrdp-web`
- Supported bpp's: 1, 16, 24, 32 (8bpp is not supported yet, palette messages handling
should be added first)
- Alpha blending is supported, and done via software-based compositing
- Inverted cursor pixels are supported (FreeRDP do not support them correctly)
- Large pointers are supported (FreeRDP crashes on them)
- Cursor caching is supported
- Add new CLI argument to enable/disable pointer rendering (--no-server-pointer)
- `session`/`client` refactoring to facilitate multiple types of session output updates besides
framebuffer update (e.g. request to hide/show system pointer on client)
- Minor changes in web client typescript code to hide/show cursor on canvas
Closes#108
Changes IronRDP web module:
- `clientName` field in Client Core Data is now configurable (but
defaults to "ironrdp-web")
- Set supported bitmap color depth to 16bpp
Base all library errors on `ironrdp_error::Error`, a lightweight and
`no_std`-compatible generic `Error` type.
A custom consumer-defined type (such as `PduErrorKind`) for
domain-specific details is wrapped by this type.
> Make the root of the workspace a virtual manifest. It might
> be tempting to put the main crate into the root, but that
> pollutes the root with src/, requires passing --workspace to
> every Cargo command, and adds an exception to an otherwise
> consistent structure.
> Don’t succumb to the temptation to strip common prefix
> from folder names. If each crate is named exactly as the
> folder it lives in, navigation and renames become easier.
> Cargo.tomls of reverse dependencies mention both the folder
> and the crate name, it’s useful when they are exactly the
> same.
Source:
https://matklad.github.io/2021/08/22/large-rust-workspaces.html#Smaller-Tips
