…ent (#5534)

## Problem

Us (QSDA) team is launching a new code generation agent.

We have conducted experiments wIth the new code generation agent, we
observed:
- Removing planning phase produces higher quality code
- Removing planning phase has general better user experience and remove
the inconsistency of code gen & plan gen

We have discussed with product that the rollout of new code generation
agent should also include planning phase removal as a whole, without an
A/B testing mechanism nor backward compatible implementation. The
endpoint for plan generation will still exist to support older toolkit
versions.

We plan to release the new code generation agent experience in the week
of September 9th.

Please squash merge this pull request, thank you!

## Solution

- Removed planning phase & refine plan in a typical /dev flow.
- Modified customer facing copies.

#5543)

…changed (#5559)

## Problem
There is a discrepancy between STE telemetry and toolkit telemetry in
terms of field `cwsprChatHasProjectContext` in
`AmazonqAddMessage`.

## Solution

Use same boolean field for `cwsprChatHasProjectContext`

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Will Lo <[email protected]>

## Problem:

The original implementation of DiskCacheError can be simplified and
still achieve it's original purpose.

## Solution:

During token refresh we eventually write to the disk cache. Sometimes
this process fails due to filesystem errors.

We dont need to care about the actual file system error, but only that
the error came during a DiskCache operation (i.e when it attempts to
save the token to disk)

So instead, at the root of where the disk cache errors ocurr, we will
wrap caught errors in a DiskCacheError.
This adds context to the error and allows anything upstream to
specifically check for that context, instead of needing to parse the
error to see if it was filesystem related.

Additionally not all file system errors were guaranteed to come from the
disk cache, so this change allows us to target the right errors.

Signed-off-by: Nikolas Komonen <[email protected]>

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Signed-off-by: Nikolas Komonen <[email protected]>

Setup Uninstall Handler for extensions.

- [x] Telemetry and Logger needs to be verified. 

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

Update Changelog for Amazon Q extesnion.
---

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

… dependencies (#5594)

## Problem

The vulnerabilities include:

```
body-parser  <1.20.3 - body-parser vulnerable to denial of service when url encoding is enabled
  express  <=4.19.2 || 5.0.0-alpha.1 - 5.0.0-beta.3

path-to-regexp  <=0.1.9 || 0.2.0 - 1.8.0 || 4.0.0 - 6.2.2 - path-to-regexp outputs backtracking regular expressions
  node_modules/@koa/router/node_modules/path-to-regexp
  node_modules/express/node_modules/path-to-regexp
  node_modules/path-to-regexp

send  <0.19.0
send vulnerable to template injection that can lead to XSS
node_modules/send
  serve-static  <=1.16.0
  Depends on vulnerable versions of send
  node_modules/serve-static
```

## Solution

Just run `npm audit fix`. It doesn't look like there's any top level
dependency changes. It looks like its all related to
[express](https://github.com/expressjs/express), which is used for the
webpack dev server

## Problem
- When you build the project amazonQSessionConfigurationMessage and
ssoCacheError get removed from the amazon q package.json

## Solution
- Add them into core, which makes sure they get copied to the
package.json of amazonq

## Problem

E2E test is not up to date post approach removal

## Solution

- Removed tests related to approach
- Refactored the E2E test script

…ange (#5581)

redo of: #3663 without
formatting problems.

## Problem
when changing instance status, the node will visually update, but only
once. Thus, starting a node will leave it stuck in pending status until
another refresh from elsewhere allows it to update.
## Solution
We generalize the solution to that used in the
`./src/apprunner/explorer/apprunnerNode.ts` into
`packages/core/src/shared/utilities/pollingSet.ts`.
- keep a set of items. 
- if the set is non-empty, run some `action` every `pollingInterval`
seconds.

In the context of ec2, this means:
- when an instance enters the `pending` status, we add it to the set. 
- Every 5, we check each node in the `pending` status to see if its
status has updated. (action)
- If the node is no longer pending, remove it from the set and refresh
its corresponding node in explorer.
- Once no more nodes are pending, we cancel the timer. Only to be
restart if a new one is added.

Notes:
- When a node leaves the pending status, we only refresh that node, not
the entire subtree.
- We check every 5 seconds if the status changed, which involves an API
call for each pending node.
- We also remove the ability to start/stop an instance that is currently
pending.
<!---
    REMINDER:
    - Read CONTRIBUTING.md first.
    - Add test coverage for your changes.
    - Update the changelog using `npm run newChange`.
    - Link to related issues/commits.
    - Testing: how did you test your changes?
    - Screenshots
-->

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

This attempts to fix our dependabot issue but I am not sure what the
best way to test this is other than committing it and seeing what
happens.

Problem (as I understand it, may be incorrect): Dependabot wants to
update deps in src.gen/, but this is managed separately and can cause
things to break. There is no exclude dirs option in dependabot. So, we
limit the dependabot to packages/core only which doesn't paint the full
picture of all of our dependencies AND since this is a monorepo where
the package-lock.json lives in the root, it cannot be updated. This
effectively bricks dependabot and makes the PRs nothing more than a
reminder.

Solution: Use this 'workaround' to exclude a directory so that
dependabot can update the deps of all of our subprojects and the root
directory without touching src.gen/.
Hack from: dependabot/dependabot-core#4364

MAY have to follow up with `versioning-strategy: 'increase'`, see
dependabot/dependabot-core#4993

## Problem
- This test started failing around sept 10th with no clear code change
- #5587

## Solution
- Until we can figure out the problem lets just disable it on mac ci

#3687)

## Problem
Potential UX Improvements. 
## Solution

![image](https://github.com/aws/aws-toolkit-vscode/assets/42325418/505b3ada-7d9c-4221-94bb-186837a9242d)

![image](https://github.com/aws/aws-toolkit-vscode/assets/42325418/76f7fac8-38eb-496a-a8db-f584b025e6b0)

<!---
    REMINDER:
    - Read CONTRIBUTING.md first.
    - Add test coverage for your changes.
    - Update the changelog using `npm run newChange`.
    - Link to related issues/commits.
    - Testing: how did you test your changes?
    - Screenshots
-->

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

---------

Co-authored-by: JadenSimon <[email protected]>
Co-authored-by: Justin M. Keyes <[email protected]>
Co-authored-by: Weinstock <[email protected]>

## Problem
Customer not able to list the provisioned clusters in region where
serverless API's not present.
Solution


## Solution
Made a minor code change to fix the issue. Tested the change too.


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Problem: Users can open multiple VSC sessions. If they sign out or the
connection is lost in another window, then interacting with chat or
in-line suggestions in the original window will only produce cryptic
"Profile does not exist" errors, with no clear way to get back to a
working state.

Solution: Log out once this state is detected (uses chat, invokes manual
or auto inline suggestion).

3 commits:

[fix(amazonq): user stays signed in if they log out in other
windows](270a321)
Problem: Users can open multiple VSC sessions. If they sign out or the
connection is lost in another window, then interacting with chat or
in-line suggestions in the original window will only produce cryptic
"Profile does not exist" errors, with no clear way to get back to a
working state.

Solution: Log out once this state is detected (uses chat, invokes manual
or auto inline suggestion).

[fix(amazonq): disable 'connection already exists' error for login
page](5593784)
Error would appear due to certain race conditions. If the auth page is
visible then we know that we have no connections. (for amazon q only).

[fix(auth): login state updates across
windows](ced109f)
- Uses a vscode filwatcher to detect changes to `.aws/sso/cache`.
- Try to 'reconnect' on new changes based on stored state key.
  - Global state is stored across windows.
- Login or log out will happen in all vsc instances.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem
New version of MynahUI requires release in vscode.

## Solution
Release new version of MynahUI with the following bug fixes:

- Fix shifted chat contents
- Fix tooltip remaining on screen after closing and opening chat panel:
#5535

MynahUI v4.15.10 release notes:
https://github.com/aws/mynah-ui/releases/tag/v4.15.10

Also fixed a regression in the following MynahUI v4.15.11
Release notes: https://github.com/aws/mynah-ui/releases/tag/v4.15.11

…tests (#5540)

## Problem
Our performance tests are occasionally flaky

## Solution
- Separate the setup, execution, and verification of the tests into different steps. Only performance test the actual thing under test
- Up the test runs to 10
- Use the median performance result rather than the average of all of them

TODO verify if trimmed mean makes more sense in this case

## Problem
- Issue with loading icon in the status bar menu.

![image](https://github.com/user-attachments/assets/94564f04-265d-49df-8a68-39745c7703b3)


## Solution
- Replaced previous icon with relevant loading icon.

![image](https://github.com/user-attachments/assets/72efac08-4a10-4cf4-9dbf-626793cc8fcb)
- Did not update any logic in the PR, its just a UI icon change so
adding/updating tests is not considered.

## Problem

In AL2, node js runtime v18 is not supported due to GLIBC <= 2.26.0.
This LSP will show user facing error message, which is disturbing.

## Solution

Disable LSP in AL2. 

In the long run, we may downgrade node 18 to 16 if possible. Meanwhile,
VS Code will soon require node v18. AL2 will migrate to AL2023 which has
newer GLIBC.

…orkspace what does test.ts do? `. (#5599)

## Problem
When filename is in the query, chat can not resolve to that file context

## Solution
Upgrade LSP to 0.1.9 to fix above use cases.

## Problem

Before:

![image](https://github.com/user-attachments/assets/6caa3a6e-3650-4d54-8eac-48484db14224)



## Solution

After:

![image](https://github.com/user-attachments/assets/e62d4090-ea0f-4a4f-bf50-f775f3c12830)

…5624)

## Problem
Gives option to copy instance id on non-instance node. 
<img width="487" alt="image"
src="https://github.com/user-attachments/assets/3a203ff8-e31c-4c85-929b-4f0fb2fbc265">
Leads to error: 
<img width="523" alt="image"
src="https://github.com/user-attachments/assets/edeb3edd-e59b-4465-8ee3-e81e6bee858d">

## Solution
Only give this option on ec2InstanceNodes

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Problem:
This job fails in the staging repo:

    Fetching the repository
      /usr/bin/git -c protocol.version=2 fetch --no-tags --prune --no-recurse-submodules --depth=20 origin …
      remote: Repository not found.
      Error: fatal: repository 'https://github.com/aws/aws-toolkit-vscode-staging/' not found
      The process '/usr/bin/git' failed with exit code 128

Solution:
Skip it in the staging repo until a better solution can be found.

see c0f3faa


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem

Post feature dev agent update, some of the texts & screenshot of the
usage are outdated

## Solution
- Updated demo gif
- Updated texts in README/marketplace page related to /dev, the text
update is reviewed by tech writer.

## Problem
Telemetry update is blocking the tracing pull request
(#5591)

## Solution
Update to the latest released version of the telemetry package

https://github.com/aws/aws-toolkit-vscode/pull/5593/files refactors how
languages are done in cwspr but its wip and this update is blocking the
tracing PR

)

## Problem

We observed a couple customer copy papercuts related to /dev

## Solution

- Updated customer facing text copies in /dev 
- Changed documentation link
- Removed a close session messaging when clicking for new task

…#5630)

## Problem
disable inlineCompletion support for language `Dart`, `SystemVerilog`,
`Lua`, `Swift`, `Vue`, `Powershell`, `R` which were added by #5621 and
we disable it because service is not ready yet thus will causing 4xx
error `improper formed request`.


## Solution
Disable Dart, SystemVerilog, Lua, Swift, Vue, Powershell, R

## Problem:

As part of our design we intentionally separate the Auth state between a
local environment, versus a remote (ssh'd) one. We do this so that if
the user has 2 IDE instances, one local and the other remote, if one
changes its auth state the other one will not be affected.

The reason for separating auth that I can remember is due to how we
store the actual tokens on disk. Because the remote does not have access
to the disk of the local, we need to ensure the remote operates its auth
independently of the local auth.

By design globalState is shared by both local and remote instances. So
for local we need to ensure we do not use the base globalState,
otherwise the same Auth state as the remote will be used.

The problem, SecondaryAuth is not doing this for all cases. It is using
the globalState when it should be asking Auth which state it should use.
The happy path was working though, but there can be potential issues
with this bug at some other point.

## Solution:

Expose the state that Auth is using so that something like SecondaryAuth
can ask Auth for the correct state object depending on the local vs
remote.


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Signed-off-by: Nikolas Komonen <[email protected]>
Signed-off-by: nkomonen-amazon <[email protected]>

## Problem
Log message can appear in output channels(such as Amazon Q Logs as
shown), IDE Debug console, or write to pre-defined log file
destinations. They come from all parts of our services and can amount to
large and lengthy chunks quickly.

We don’t have a reliable way to identify log messages from various
services and modules of our product. Some messages might have headers
while others don’t, and the existing headers are hard-coded on a message
by message basis, with no checks for consistency.

1. It’s difficult to locate and trace unidentified log messages when the
size of logs grow quickly.
2. Without regulation, the hard-coded custom headers can lead to legacy
issues down-the-road and inconsistencies within the codebase.

## Solution
1. Make message topic identifiers an input variable of the getLogger()
function.
2. Append the topic at the front of each message, for example:
`getLogger('S3').error('Something went wrong')`
Would print:
`S3: Something went wrong`
3. Define a set of topics to be used in message headers, covering all
services and modules.


## Implementation notes:
1. The `WinstonToolkitLogger` is renamed to `ToolkitLogger` to avoid
confusion, since its our own logger not winston's. The log topic feature
is added here.
2. A `baseLogger` abstract class is created to make the existing
`NullLogger`, `ConsoleLogger`, and `ToolkitLogger` cleaner and their
responsibilities more obvious and compact.
3. Currently the topics are optional, and the values are just `Test` and
`Unknown`, this will be expanded during migration of existing logger
calls.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Nikolas Komonen <[email protected]>

## Problem
- We can't see an e2e view of our events

## Solution
- Implement telemetry tracing using the opentelemetry idea of trace ids.
- A trace id helps create a collection of spans.
- Only one trace id can be active in an execution flow at a time.
- Additionally, it swaps parentMetric for parentId so that we can
directly correlate one telemetry event to another

## Other details
- A trace is associated with a top level telemetry event. The top level
can be either event_trace or any other of our metrics.

## Problem 
We currently check if certain policies are attached to the IAM role in
order to determine if the necessary permissions exist on the instance to
open an SSM session. However, this can be done more fine-grained.

## Solution 

https://github.com/aws/aws-toolkit-vscode/blob/a9d64f3a448480f8fed9ba96b3da33631c298265/src/ecs/util.ts#L33-L52
Leverage the approach done by ECS to check for specific actions rather
than policies.

- Restructure testing framework to look at actions instead of policies. 
- Refactor ECS code to pull out shared functionality to
`src/shared/remoteSession.ts`.
- Refactor existing code in the `src/ec2/model.ts` file to utilize the
approach demonstrated above.
- Add back `packages/core/src/shared/systemUtilities.ts` file since some
of it is used here.
<!---
    REMINDER:
    - Read CONTRIBUTING.md first.
    - Add test coverage for your changes.
    - Update the changelog using `npm run newChange`.
    - Link to related issues/commits.
    - Testing: how did you test your changes?
    - Screenshots
-->

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

---------

Co-authored-by: JadenSimon <[email protected]>
Co-authored-by: Justin M. Keyes <[email protected]>
Co-authored-by: Weinstock <[email protected]>

## Problem
The IDE force snaps to Debug Console during local lambda build

## Solution
Create `showOutputChannel` function to focus the window on output
channel

## Problem
User Request
## Solution
Add option to launch instance from toolkit. The option links to console
page for launching EC2 instances. Currently, the functionality is
available as an icon-button and on right-click. Additionally, we have it
in the command-palette. In the command palette, the region is inferred
using `guessDefaultRegion`. I see three potential alternatives:
- (1)provide a region prompter when used from the command palette that
defaults to the region given by `guessDefaultRegion`.
- (2)Don't prompt for region, and assume the region based on
`guessDefaultRegion`.
- (3)Don't allow the command to be used from command palette and only
see it on the explorer.
Let me know if you think either of the alternatives presented above
would be preferred.

The current implementation (1) because it seems like the optimal middle
ground. We allow the customer to select their region, while also
providing an intuitive and helpful default.

Below are some screenshots of the feature:
<img width="743" alt="image"
src="https://github.com/user-attachments/assets/3f56800c-ca95-4f82-be48-76d353ad92ad">
<img width="530" alt="image"
src="https://github.com/user-attachments/assets/088da856-7bbc-4e3f-98bf-1028372111bf">
<img width="691" alt="image"
src="https://github.com/user-attachments/assets/a09367d2-cd41-4f6a-804f-fb7327041f53">


<!---
    REMINDER:
    - Read CONTRIBUTING.md first.
    - Add test coverage for your changes.
    - Update the changelog using `npm run newChange`.
    - Link to related issues/commits.
    - Testing: how did you test your changes?
    - Screenshots
-->

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

---------

Co-authored-by: JadenSimon <[email protected]>
Co-authored-by: Weinstock <[email protected]>
Co-authored-by: Justin M. Keyes <[email protected]>

## Problem
Occasionally this error pops up: 

![image](https://github.com/user-attachments/assets/28630fd3-569e-4307-914d-815669686ba4)

## Solution
Add one line to ssh config that force agent to use the provided key. 
- `IdentitiesOnly yes`: only use identity file provided. 

More info here: https://www.ssh.com/academy/ssh/config

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Justin M. Keyes <[email protected]>

## Problem
currently use RSA.

## Solution
switch to ed25519

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Justin M. Keyes <[email protected]>

## Problem
By default, the ec2 instances do not have the proper permissions to
allow for connect via SSM. Therefore, users will have to complete this
step manually.
## Current Solution
Check if IAM role exists on instance, if it does, prompt to add
necessary actions as inline policy.
### Details and Problems with Current Solution
- actions in prompt are poorly formatted. 
- attaching the inline policy, then immediately attempting connection
does not appear to work reliably. The cause of this is currently
unknown, but one potential explanation is the delay of propagating
permissions across all regions. When immediately attempting connection
after adding the inline policy, the terminal will open, but will
immediately crash. Note that no error is thrown by the SDK here, meaning
we have no way of knowing if the terminal will crash or not before
opening the terminal. Additionally, while we do get a terminal close
event on the crash, we have no way of distinguishing this from a user
close event. The current solution to this problem is to add a 5 second
delay between adding the policies and attempting connection. During this
process we provide a progress bar with cancel. This solution is far from
ideal, but does appear to work. A polling approach was also attempted,
but seemed to fail as the SDK calls returned that the necessary policies
were attached before a connection can be successfully made.
- after attaching the inline policy, there is often a delay in the SSM
agent starting on the instance. This results in receiving the `is SSM
agent running error?` if connection is attempted immediately following
adding the inline-policy (only in remote window, this error doesn't
happen when trying to establish a terminal). This can take up to a few
minutes to go away, or can be immediately resolved by fully restarting
(not rebooting) the instance. Giving the customer an option to restart
is not a feasible solution as it comes with many side effects (new
billing cycle, new ip address, etc.).
- Actions added include 
```
    'ssmmessages:CreateControlChannel',
    'ssmmessages:CreateDataChannel',
    'ssmmessages:OpenControlChannel',
    'ssmmessages:OpenDataChannel',
    'ssm:DescribeAssociation',
    'ssm:ListAssociations',
    'ssm:UpdateInstanceInformation'
```
- The bottom three, while not necessary for the connection, appear to
help in starting the SSM agent on initial connection. With these
policies attached, and restarting the instance, the connection works
every time.
- Actions added are expanded such that SSM agent starts on instance. 
- If no IAM role attached, we link to documentation. 

### Visuals

progress bar with cancel:
<img width="539" alt="image"
src="https://github.com/aws/aws-toolkit-vscode/assets/42325418/b144ed2e-5bf2-41a2-b7f2-0f56e598ea10">

prompt with actions: 
<img width="306" alt="image"
src="https://github.com/user-attachments/assets/c9c0aa1c-8e13-4c4d-a178-f503e633647d">


<!---
    REMINDER:
    - Read CONTRIBUTING.md first.
    - Add test coverage for your changes.
    - Update the changelog using `npm run newChange`.
    - Link to related issues/commits.
    - Testing: how did you test your changes?
    - Screenshots
-->

## License

By submitting this pull request, I confirm that my contribution is made
under the terms of the Apache 2.0 license.

---------

Co-authored-by: JadenSimon <[email protected]>
Co-authored-by: Justin M. Keyes <[email protected]>
Co-authored-by: Weinstock <[email protected]>

## Problem

For connecting VSCode to EC2 instance, we generate an ssh key pair on
disk. This results in writing the private ssh key to VSCode global
storage, allowing the key to be potentially reused by other users on the
same machine.

## Solution  

Attach a lifetime to any key pair generated such that they wipe from
disk after X seconds. Value is currently set is 30 seconds to allow
connection to reliably establish. Also, change file permissions to
read/write owner only and change behavior to overwrite existing keys.

Unable to test file permissions, due to VSCode file system unable to
provide us with enough detail. Only provides whether it is readonly or
not (somewhat unreliably:
microsoft/vscode-discussions#673). VSCode
file system is what is used in `fs.ts` implementation.

This is in-line with how ec2 instance connect works:
https://github.com/aws/aws-ec2-instance-connect-cli/blob/master/ec2instanceconnectcli/EC2InstanceConnectKey.py


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: JadenSimon <[email protected]>
Co-authored-by: Justin M. Keyes <[email protected]>
Co-authored-by: Weinstock <[email protected]>

## Problem
No documentation about traceIds

## Solution
Add documentation about traceIds

## Problem
- We have no way to connect disjoint events that may or may not be in a
telemetry span


## Solution
- Add withTraceId, which allows you to start a new trace with a specific
id
- Update the telemetry docs

## Problem
Hardcoded credentials detections are being removed from auto scan
results due to line string comparison with the server response which
contains redacted code.


## Solution
For redacted code, check only the line length instead of a full
comparison.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem
Some users transform projects with long-running tests.

## Solution
Allow users to skip tests by prompting them via chat and passing the
relevant flag to skip tests through the `manifest.json`.

## Problem
- complicated regex which is not scalable and hard to understand
- stale and complicated code which lacks readability
- test coverage

## Problem
(title)

## Solution
Try to use `ed25519` first, if it fails use `rsa`. Alternatively, we
could use the same code to check if version number of ssh, and then see
if its above version 6.5
(https://www.openssh.com/releasenotes.html#6.5). However, doing an
actual attempt at generating the keys felt more robust, as it could
potentially fail for other reasons. Make use of the code here:

https://github.com/aws/aws-toolkit-vscode/blob/791ae55021c0ee73dd84abaa57da46237d0fb0d5/packages/core/src/shared/utilities/pathFind.ts#L117-L134

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Another PR I have adds a process related function, so I thought I'd make
the name more appropriate.

This is a separate PR since it adds lots of noise to the other PR if I
do it there

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Signed-off-by: nkomonen-amazon <[email protected]>

## Problem


## Solution


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem
- The flow from when a message is shown to the user and
rejected/accepted by the user isn't instrumented with any way to trace
events

## Solution
- Create a trace from when the event is shown (either by manual
invocation or auto trigger) and then connect all the telemetry events
through a traceId.

### Implementation details
- The traceId is initially created when manual invocation/auto trigger
occurs
- For every telemetry event thats in the function hierarchy of the
manual invocation/auto trigger, they automatically assume the same
traceId
- When an event listener is called it uses the traceId we've set in
TelemetryHelper to setup the next traceId. This is because the event no
longer belongs to the same context
- Inside of commands we use `telemetry.record` because it allows us to
override the default telemetry id that gets automatically created and
instead associate it with the known trace

…tions (#5654)

## Problem
Need to gather acceptance metrics on generated unit tests

## Solution
Include user intent on copy and insert code interactions in telemetry to
form an acceptance rate

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem:
- We can't measure e2e how long a chat request takes

## Solution:
- Plumb the trace id through to the webview -> vscode -> webview. This
allows us to know exactly what is happening to a message and how long
everything took to get there
- chat e2e latency can be linked with the other telemetry events via the
traceId

This adds some util functions that are needed for an upcoming feature.

Each commit is its own isolated util, so review them individually.

**A significant amount of the changes come from minor refactors in
files**

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Signed-off-by: nkomonen-amazon <[email protected]>

verify path for .obsolete file


License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Problem:
Redundant code, suboptimal names and docstrings.

Solution:
- Introduce `toTextDocument` for tests such as `codeParsingUtil.test.ts`
  which don't need a TextEditor.
- Rename `openATextEditorWithText` => `toTextEditor`.

- Add command to Amazon Q to generate unit tests for selected code. The
  feature will be limited to internal Amazon users, initially.
- Add "Generate Tests" command.

… for all users (#5674)

## Problem

Unit test generation feature is not ready for internal users yet. This
PR disables the feature while improvements are made.

## Problem

Revert #5607 by not
showing users the skip tests form


## Solution

Above

Problem:
Regression b2ebc10 in the logger: `debug('%s %s', foo, bar')`
sends `foo, bar` as a single `[foo, bar]` list to the first `%s`.
Example:

2024-09-26 12:18:29.631 [debug] getClientId: determined clientId as: [
      '96be923b-a00d-46d4-add1-c877abfcb01c',
      undefined,
      '96be923b-a00d-46d4-add1-c877abfcb01c'
    ], process.env clientId was: { logID: 1 }, stored clientId was: %s

Solution:
Spread the args when passing to sendToLog().

## Problem
In VS Code, when user add 2 different folders in the workspace which
contain different root paths, RTS/LLM cannot decide where to place a
generated file.

## Solution
This fix add the default first folder selected as its default for
orphaned root created files, not blocking user to receive an empty patch
but also support better the workspace usage.

Theory: dependabot is still trying to update src.gen/ dependencies
because they are managed by the root monorepo. So technically, there is
no way to avoid updating src.gen's package.json because the root
package.json is linked to it and impacted by dependabot changes for
subproject dependencies.

Let's make its own "subrepo" and see if we can get dependabot to ignore
it finally.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Removes the need to copy/sync package JSON contents from core ->
subproject extensions on build and package.

Summary of changes:
- Delete syncPackageJson.ts and handlePackageJson.ts
- Consolidate testing scripts and make small improvements (e.g.
validation, handle multiple directories to import tests from)
- package.json is now fully complete in `amazonq/`, `toolkit/`. Changes
to extensions should be made only the respective file (except for
icons).
- `toolkit/` runs unit, e2e, and integ tests from `core/`. It will also
run any of its own tests (but those don't exist yet).
- Web tests now moved to and run from `amazonq/`. This was decided
because `amazonq/` is more likely to support web than `toolkit/` at this
time. Also, it didn't seem possible to run the web tests from `core/`
like the other tests because it relies on webpack and a separate test
runner.
- No longer create web-based webpacks for `core/`

Follow up work:
- Get [packageJSON
properties](https://github.com/aws/aws-toolkit-vscode/blob/46b409c3353ce0dafefb796f5994c1fa88a23b47/packages/core/src/shared/vscode/env.ts#L61-L74)
from `context.extension.packageJSON` instead.
- Cleanup any remaining unneeded tasks in `.vscode/tasks.json`
- Move icons out of `core/package.json`.
- Continue to investigate: can we run web tests from core like other
tests? Toolkit may have a use in web mode one day.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

… item (#5699)

## Problem

Command and context menu item for unit test generation was disabled for
internal users before last release. We are ready to enable now.

## Solution

This PR enables the context menu item and commands for unit test
generation.

## Problem
The test case here:
https://github.com/aws/aws-toolkit-vscode/blob/359b1d8252a32e064482678110c5931630bd49e0/packages/core/src/test/awsService/ec2/model.test.ts#L132C9-L146C11

This test is unaware of the side-effect that it will attempt to
explicitly delete the keys when generating. Originally, the delete was
conditional on the keys existing in the directory, so this test case was
not a problem since they never existed. However, on windows that check
was faulty, leading to the case where the windows file system reported
that the keys existed, which lead it to attempting to delete the root
directory.

## Solution
- Modify `tryRun` to accept an `onStdOut` optional parameter that lets
us accept the overwrite prompt from `ssh-keygen`.
- Do not explicitly delete the files in the generation function, make
this part of the generation step.

The benefits of this approach are that we moved the side-effects to a
place where they are expected, in the key generation process itself. It
also inherently avoids the potentially faulty check if the key files
exist, as we only use `onStdOut` if when the keys exist.

### Notes
- `packages/core/src/awsService/ec2/sshKeyPair.ts:49` includes `any`
which should be `ChildProcess`. For some reason the import fails in CI,
so I left it as `any`. Refer to `attempt to import child process` commit
for logs.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Justin M. Keyes <[email protected]>

## Problem
The issue is here:
https://github.com/aws/aws-toolkit-vscode/blob/9eea1f3700930df79620ef26ff6a43450e593580/packages/core/src/awsService/ec2/explorer/ec2ParentNode.ts#L55-L64
This fires off multiple asynchronous calls on each child, with the
function returning before any are resolved. Thus, the child node could
be deleted from the set after this function returned (potentially when
another piece of code is using it). This leads to `undefined` errors
within that function.
## Solution
Refactor the code to make use of a `for...of` with `await` pattern. 

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem
Changing the permissions of a file is currently not supported by the
file system wrapper.

## Solution
Implement a stub chmod that does nothing on web, and uses `fs` to modify
permissions in non-web environments.

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

---------

Co-authored-by: Justin M. Keyes <[email protected]>

## Problem
We will rename Application Composer to Infrastructure Composer as our
service name. We need to do the same thing in VSCode toolkit

## Solution
We will rename all user visible things first

make auth container responsive

# Before - The content scales down to 0.6x when the width in less than
260px

https://github.com/user-attachments/assets/d38db517-4734-435f-8100-b6a65d502c75

# After
## Login Screen

https://github.com/user-attachments/assets/17b579e1-003f-42aa-967e-7565e66624f0

## Re-Auth Screen

https://github.com/user-attachments/assets/2cd8754d-6707-4cd4-a250-7b28a0cd7de5



License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Description
Updated the legal disclaimer text in the footer of the chat client. This
is done as a clarification of compliant use according to the AWS AI
policy.

We dont need this code anymore

---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Signed-off-by: nkomonen-amazon <[email protected]>

## Problem
- withPerformance was an experiment to see if we could correlate
cpu/memory with number of files


## Solution
- Cpu usage/memory usage in the extension host is way too volitile so we
might as well clean up this code


---

<!--- REMINDER: Ensure that your PR meets the guidelines in
CONTRIBUTING.md -->

License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

## Problem
IamPolicyChecks currently has no unit tests. This makes the feature
fragile to changes and at risk to breaking changes.

## Solution
Implement unit tests for IamPolicyChecks feature.