feat: implement adb pull for usb

[lavafroth]

Changes

• Added USBSubcommand enum for subcommands like List, Stat, etc.
• Added private send_and_expect_okay and recv_and_reply_okay method to ADBUSBDevice since the host and the device quite frequently acknowledge each others' messages with OKAY commands.
• New members in ADBUSBDeviceExt trait
  • A stat method to stat a file on the device.
  • The pull function which takes a source file path (string) to read from the device and writes its contents to a Writer output.
  • Added USB subcommands pull to the CLI

Note

stat and pull are implemented for the USB side but they are left as todo!s for ADBServerDevice

To pull a file called src from the device and save it as dst, run

cargo r -- usb --vendor-id 2209 --product-id 2700 pull src dst

[cocool97]

Thanks for the PR !

• Can you add the code in adb_cli ? You can see how it's done in pull for physical devices
• Can you add at least the pull function in the trait ? It is implemented in physical devices and just needs to be moved :)

I'll review your PR and use some functions you wrote for the main USB branch

[lavafroth]

Hello, I added a subcommand to adb_cli now you can run the command as described above to pull a file and save it to a destination. I'm afraid I didn't quite understand your second point. I have added the pull function in the trait. It is just not implemented for the ADBServerDevice yet. 😅

[cocool97]

Pull command is implemented, just named recv 😉
I think we can rename it to be more consistent with upstream ADB cli !

[lavafroth]

Okay thank you. I have moved the recv method to the ext trait. The server, as you mentioned, uses the existing implementation. The ADBUsbDevice uses the implementation from this pr.

[cocool97]

Thanks for your improvement. There are still some things that might be improved :

• Stat method in ADBDeviceExt needs a local_id and remote_id, but some other methods also need it without require it in public API (like shell e.g). From a user perspective, these identifiers should be hidden. Must an OPEN command be sent before trying to pull a file from device ? If yes, we can merge behaviours with shell and maybe store these identifiers in the USBTransport structure ?
• Stat is also implemented for ADBServerDevice, can you replace the todo!() with the current implementation ?
• The server implementation returns an ADBStatResponse (src/server/device_commands/stat.rs), can you merge this implementation with your FileStat to have only one ?

I've done the rebase on usb branch so after these modifications I'll test it and merge :)

[lavafroth]

Sure thing, I'll make the changes by tomorrow!

[lavafroth]

Hey there, I have made the necessary changes. The user should no longer have to be concerned about the local_id and remote_id. The server and usb stat response structs are merged as well!

[cocool97]

I had a look at your code, everything is fine. The begin_transaction method is nice, I now use it in shell too :)

Just have a minor issue before merging it, stat command seems to not return values in the correct order and does not close correctly (as seen after)

(/data/local/tmp/bigfile has just been created in this case with dd if=/dev/urandom of=/tmp/bigfile bs=100k count=1024 and has a size of 100Mb)

First run:

RUST_LOG=adb_client=debug cargo run -- usb --vendor-id 04e8 --product-id 6860 stat /data/local/tmp/bigfile
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.15s
     Running `target/debug/adb_cli usb --vendor-id 04e8 --product-id 6860 stat /data/local/tmp/bigfile`
[2024-10-25T08:15:19Z INFO  adb_client::usb::adb_usb_device] Authentication OK, device info device::ro.product.name=gts210vewifixx;ro.product.model=SM-T813;ro.product.device=gts210vewifi;features=cmd,shell_v2
File permissions: 1413567571
File size: 33279 bytes
Modification time: 1973-04-28 15:06:40.000000000 UTC

Second run:

Error: ADB request failed - Wrong command received CLSE

[cocool97]

Behaviour is the same when trying to pull the same file twice without unplugging the device

[lavafroth]

The pull and stat commands were leaving ADB on the device hanging but this has been fixed! We had to end a sync transaction using a quit command.

[cocool97]

Thanks, this works better !

Last point about stat command, output is not as expected for a single file.
To reproduce create a file, run CLI in "ADBServerDevice" mode and run it in "USB" mode, outputs are not the same. I tried and "server" mode outputs are correct.

[lavafroth]

Ah yes, it looks like there is an extra u32 in the beginning of the stat response structure. Since we are not using it, I have made the parser skip it as well.

[cocool97]

Yes, extra u32 is the litteral "STAT". I add a comment to your last commit and merge it right after

[lavafroth]

Thanks for the heads up. Let me know if I can help with implementing any other feature.