9.0 Shipped on Linux so we should update the tests. These tests do test the live availability of what's available and it's good that they do so they actually test the functionality and not some mock data, which wouldnt be a relevant test. That does mean they need to get updated, but I think the messaging is pretty clear and the impacts aren't so bad.
* Fix acquisition error on non fully qualified version
The new, unreleased code for getting a feature band of a sdk version try catches when it tries to get the band. It will not throw if there is no band in the string which can be expected in certain cases, but it still posts an error to the event stream. Posting an error to the event stream causes the error to bubble up to the user even if it is caught. We need to take some additional action here so the user doesn't see an error popup for something that's not an error. This is blocking our release with all of the findPath features.
Resolves https://github.com/dotnet/vscode-dotnet-runtime/issues/2015
Version updates are to update the shas of the internal packages which got updated.
* Update imports
* remove extra else
* Don't use setting in SDK PATH lookup
* Add worst case scenario no which or where on path
* Add test for finding sdk path and finding path setting on find path
* Fix test
* Respond to linter
* Fix typo
* Fix the test
* Consider Runtime Patch Versions in PATH API
The C# extension will now fail if the runtime is not 8.0.10 or higher on mac so we need to support patch version lookup.
Sadly we cannot migrate to semver because semver rejects strings such as 8.0 which are allowable in many parts of our code and APIs.
This also adds code for SDK lookup but no test since that's harder to test, like remarked in the code.
* Fix test
* Fix the version logic to not only parse major.minor version
We now parse all of the version, dont need to do this logic.
* Fix Version Parsing for FindPath API
The old comparison used string comparison which is fine for versions of dotnet until dotnet 10. '10.0' < '9.0'. 😂 This fixes that. I suppose it's a bit clunky and we could dedupe / share the compare logic but it's not being used elsewhere in this same form, so I don't plan to do that right now. There are also places that depend on the output versions being strings, so that's a problem to change that now too. It probably should have been a class from the beginning but unfortunately that just isn't how it was initially done, which is understandable because the scope at that time was much more limited.
* Fix some of the code
* Fix test
* Dont comment out the other tests
* update comment
* Fix incorrect PATH echo
* Add Architecture Check into Find PATH API
We had a discussion about how Mac users tend to need the arm64 runtime and often have the x64 host installed on their PATH. This was a big concern. With global installs at least we can still rely on the output from `dotnet --info` for now, so we decided to add this back in.
* Fix test, the runtime install does not have an arch printed so we cant quite use it to check properly
* Fix test
* Fix test
* undo comment out
* Fix a very terrible bug due to a typo
* code cleanup
* Try asking the user if the want to continue
* Add a message on mac if someone tries to cancel the installation
Consider that to tell if the user wanted to say 'no' or 'yes' to retry, that doesnt indicate whether they meant to do the install or not. We would then need more options to tell if we actually failed or not. I dont want to rely on this and I dont think it is a high enough priority with the complexity required to reroute the code to add a retry button that does a certain retry depending on if someone clicked to cancel the install on mac
* Consider if the dotnet file doesnt exist but the folder does
* Fix bug
Certain events get sent way too often (ex: 700 million locks acquired over a few months which is causing problems for the data team) and aren't useful for telemetry.
We allow suppressing certain events before even hitting the data ingestion service by doing a check here.
I made a change a while back such that 'cancelled request' style errors dont cause a pop up because I figured they were not useful. The vendors reported it as a bug. Do you think we should keep a pop up for those types of errors?
The types of errors include: entered a bad version, you cancelled the password prompt or admin prompt, you have a custom linux install or conflicting feeds, your distro is not supported, etc
Resolves https://github.com/dotnet/vscode-dotnet-runtime/issues/1826
If this is closed, close this issue.
Waiting for PM feedback on @baronfel. I think we can display it.
Extensions can set errorConfiguration in the acquire context
export declare enum AcquireErrorConfiguration {
DisplayAllErrorPopups = 0,
DisableErrorPopups = 1
}
to ignore this.
* Add back output for uninstall in the terminal
Resolves https://github.com/dotnet/vscode-dotnet-runtime/issues/1920
There used to be this code to print a message to the terminal when an uninstall is going on. It got deleted at some point.
* Clean up import
By passing recursive:true to all mkdirSync calls we protected against
issues that may arise from parent directories not already existing.
Fix#1985
Co-authored-by: Andy Sleigh <andy.sleigh@healthhero.com>
* Permit Installing .NET via a package.json file
Extensions that rely on .NET often have a slow startup time due to our existing API.
For extensions that want the .NET runtime to be installed as soon as possible, they can now also make an API request in their package.json.
They should add a [IDotnetAcquireContext](https://github.com/dotnet/vscode-dotnet-runtime/blob/main/vscode-dotnet-runtime-library/src/IDotnetAcquireContext.ts) object in a section titled 'x-dotnet-acquire' to do so. They don't need to include the `requestingExtensionId`, since this is already in their `package.json`,
This should go at the root of a `package.json`, and not in the `contributes` section.
```
"x-dotnet-acquire": {
"version": "8.0",
"mode": "aspnetcore"
}
```
We run the install on startup and also whenever a new extension is installed via the onDidChange API. https://code.visualstudio.com/api/references/vscode-api#Extension<T>
You can easily test this by modifying the SDK extension and adding this, then packging t with `vsce`, installing it, and running the Runtime Extension under the Debug tab. If you do so, you will see that:
If you install a new extension, the request is made.
If you had already installed the extension, the request is also made.
* Respond to linter
* Add a very basic test.
* Fix typo
* Fix typo
* update PATH setting description to make it more clear
* add very prototype code
* add tests and get runtime check working
* Fix some bugs
* Consider that the .NET SDK can also satisfy the Runtime
* fix typo
* fix the tests
* fix
* fix another test
* Fix tests even more
* undo yarn changes
* fix the final test hopefully
* undo yarn changes
* Dont use dotnet.exe as its not platform agnostic
* use "markdownDescription" for nicer rendering in VSCode
* Tweak call to action to use VSCode mechanisms first, then our standard installation docs, and only in the last resort mention PATH munging.
* Fix URLs
* Move to version 2.1.7
* add basic command'
* Ignore existingPath setting for SDK installs.
C# DevKit never uses the path returned by our installation. This means users would think this path would change the sdk that this extension uses but that is not the case. This path to dotnet.exe is meant to be the path for the runtime for extensions to run on, and not the SDK path. It's confusing that the setting was used for both and a misstep in a way. DevKit is the main caller of this API so we think we can change this with minimal breakage.
* Remove warning setting and fix invalid path setting
The setting must be accessed earlier. This means vscode will need to be restarted. We also update the readme and messaging a bit so its more publicly clear in all places what the setting is for.
* Fix test
* Refactor code out into a Validator for Conditions
dotnet --list-runtimes and more need to be called in more places. This is a separate task so it should be done. I did not change the code in any way except for adding the requirement clause type.
* Prepare code to validate the path
* add a lot of prototypey code
* add comment for future work
* merge with main
* look up the architecture
* Improve the code
* Go 2 Directories Up to find the True Path on PATH
* Final initial loop of API code
* Fix bug parsing list runtimes
* Add tests
* Fix test and search for where if its not installed
* Consider where may return multiple values
* Fix test
* tests mostly working
* code cleanup - get rid of extra api to set env
* Restore the env var so we dont edit it for other processes
* Uncomment the remaining tests
* Respond to lint
* fix callback
* Fix path to be os-gnostic
* Only search for where on windows and also search for which
* provide env to the find command so /usr/bin/whcih can be used
* Call which which instead of which so the correct command can be found
* Install 3.1 instead of 7.0 because the DTL CI machines seem to have a 7.0 SDK on them :zany:
* give up on arch check for now because it is inaccurate, see comment
* Add github issue in comment for context
* make linter happy
* Respond to PR feedback
* Migrate to connection strings
Resolves https://github.com/dotnet/vscode-dotnet-runtime/issues/1958
The old application insights key was created by @LakshanF, when we migrated to the new vscode-extension-telemetry service, their API had a breaking change to require a connection string instead of an insights key. https://github.com/dotnet/vscode-dotnet-runtime/pull/1948
The connection key can be public and is hard coded. Our existing key has been in our open source, source code for many years. Here is what their guidance says: https://www.npmjs.com/package/@vscode/extension-telemetry
> Follow guide to set up Application Insights in Azure and get your connection string. Don't worry about hardcoding it, it is not sensitive.
* Respond to PR Feedback
---------
Co-authored-by: Chet Husk <chusk3@gmail.com>
TS Lint failed this `Unexpected `await` of a non-Promise (non-"Thenable") value`.
But the callback can be a promise type. Its just not smart enough to realize that.
* update PATH setting description to make it more clear
* add very prototype code
* add tests and get runtime check working
* Fix some bugs
* Consider that the .NET SDK can also satisfy the Runtime
* fix typo
* fix the tests
* fix
* fix another test
* Fix tests even more
* undo yarn changes
* fix the final test hopefully
* undo yarn changes
* Dont use dotnet.exe as its not platform agnostic
* use "markdownDescription" for nicer rendering in VSCode
* Tweak call to action to use VSCode mechanisms first, then our standard installation docs, and only in the last resort mention PATH munging.
* Fix URLs
* Ignore existingPath setting for SDK installs.
C# DevKit never uses the path returned by our installation. This means users would think this path would change the sdk that this extension uses but that is not the case. This path to dotnet.exe is meant to be the path for the runtime for extensions to run on, and not the SDK path. It's confusing that the setting was used for both and a misstep in a way. DevKit is the main caller of this API so we think we can change this with minimal breakage.
* Remove warning setting and fix invalid path setting
The setting must be accessed earlier. This means vscode will need to be restarted. We also update the readme and messaging a bit so its more publicly clear in all places what the setting is for.
* Fix test
---------
Co-authored-by: Chet Husk <chusk3@gmail.com>
* upgrade to the newest version of extension telemetry
* remove vscode-extension-telemetry where its not needed
* Update vscode-test as well since its a new library
* 2.1.6 branding
* add other potential mac path from installer/generatePKG.targets -- need to add tests
* fix tests
* Simplfiy expression logic
Co-authored-by: Michael Yanni <MiYanni@microsoft.com>
* Fix syntax
---------
Co-authored-by: Michael Yanni <MiYanni@microsoft.com>
* get eslint running but its failing a lot of checks
* Fix the linting after migration
* Fix many errors
* 65 errors to go
* fix almost every lint issue
* clean lint
* Remove references to tslint
* dont install eslint as it is already here
* install eslint
* update eslint
* dont empty cache because eslint is failing to register the types that exist
* try to help eslint find the files for type defs
* Set typescript resolver
* Fix imports
* Add some import plugin but still cant import vscode correctly
* install eslint globally
* skip npm installl
* add imports
* try to fix the import issue
* build before hand so the types are picked up -_____-
* Revert "build before hand so the types are picked up -_____-"
This reverts commit e5381b42df.
* rollback to eslint 8 but still run build
* Encase the installer path in quotes to protect against spaces in username being parsed poorly, and other path issues. Don't know why we didnt do this before.
* fix linter
Rejecting a promise with a rejected promise as a return value from a function does not reject with the object the inner promise rejected with. Instead it rejects with an undefined value.
This means if the SDK installer fails or we fail to elevate, that counts as .NET Install failed: undefined. This fixes this. It also adds a check for the string error as this seems to happen sometimes if the password prompt is rejected.
How to test this? Set a break point in win mac global installer on the execute install and edit the command to point to a non existent file. It will then fail and you can see it works now.
* Fix an issue parsing promises
We changed the function `getPromise` to not use a lock in version 2.1.2. This was a mistake. I thought it would return the unresolved `Promise` object if it was a sync function. That is not the behavior, instead it will skip that line of code and return null even if the promise exists. This causes multiple processes to try to grab the installer file handle.
When one fails to grab the file handle, it causes a chain-reaction of cascading failures for concurrent requests, because that is supposed to be blocked from happening.
In addition, the check I added to kill the sudo process once it is finished runs on all platforms despite the master sudo process only being present on Linux, which has resulted in some higher timeout rates. We need to not do this on other platforms.
* Fix tests
we need to clear the promise out after each test is done
* dont call uninstall all in sdk test, its the wrong command
* reset event stream in the test
* Add failure handling for file integrity check
This is now what is hitting the EPERM and ENOENT errors.
We need to skip the check and see what happens from here, such as allowing to elevate via windows.
This also adds specific handlers for when we fail to download the SDK.
* elevate on windows
* Retry if no permission is available the first time
Need to double check that cancelling is handled properly now that its a promise reject
* respond to linter
* Add API to uninstall any install
* Add tests
* undo bad save
* dont uninstall if there are multiple owners
* add some basic uninstall capability
* allow elevation on mac
* fix promise chain a bit, may need to await still
* add uninstall functionality to linux
* add uninstall on linux
* fix merge
* Add a test
* add ui element to pick uninstall
we need to fix the uninstallglobal functoin logic to print out and such still
then test it
* add some output, need to test and check graveyard
* add uninstall events
* fix build
* Fix uninstall to pop up above window
* ui changes per request
* fix test
* fix merge err
* fix mac test
* fix name option
* fix promise logic
* fix promise code again xd
* Add icons for the extension
* add icons to shield and uninstall
* set minimum vscode version to require new icons
* dont sort if no install exists
* use noninteractive frontend in master process to fix dpk unable to re open stdin on uninstall
* try to exit sudo process master and set env var for stdin uninstall fix
* fix eventstream
* fix detection logic for uninstall
* add -y to uninstall
* fix bug with uninstall
* add version check for when disk is full and version string ui fails
* also kill sudo process on uninstall
* use -f to delete file as they may not exist and we dont want to fail due to this at the end
* add -y to update
* respond to linter
* remove unnecessary img
* respond to linter
* respond to pr feedback
* fix when extension may have null record
* add function to call in extensoin.ts for offline mode acceptAlternativeVersions on acquirectx
* refactor version helper functions outside of version resolver
version resolver is a class that is based around using the internet. It probably should have been like this before.
* add code to check existing installing versions
* Update test name
* respond to linter
* Update lint
* Fix bugs and add messages
* respond to pr feedback
* Add fake MSBuild project to allow signing VSIX and JS
* Sign JS
* Add step to pipeline
* Restrict signing package version
* Add scripts to install signing tool locally
* only go up 1 directory
* fix display name
* fix whitespace -__-
* add build step for installing microbuild
* add sign type parameter
* dont produce alpha package in official drop
* add sign type parameter to default pipeline
* dont publish package in another directory as this causes other issues, instead copy the artifact
* acknowledge the existence of a document explaining the release and other maintenance processes
this information is more suited for internal repo changes.
* change internal stream to a different signing stream
* add dotnet public for notargets sdk
* reduce to 1 feed
* use a different feed
* Use dotnet team
* use explicit feed source
* add feed auth
* fix comment
* rename unsigned vsix so both can get dropped
* use the other type of slash for REN -__-
* try to fix path
* add packages folder with empty gitignore
* figure out dir
* fix rename command
* Fix rename
the 'unsigned' one is actually the signed one
* Fix JS File Sign
dist is at the root of each but this runs in a segregated folder
* produce binlogs and try to fix the path
* Publish SDK Extension Logs
* fix signing
* remove bad slash in path that gets parsed incorrectly
* remove \
* fix microbuild to be installed b5 js signing
* update gitignore
* fix gitignore again
* [REVERT THIS] Add a file A.ts to see if files are signed alphabetical or only non minified JS is signed
* [REVERT THIS] Add content to js to see if minify changes signing
* sign js after webpack
maybe we can sign it after the bundle is created and it will still be able to edit the bundle vsix internals? I thought not but perhaps extension.js is getting replaced by the webpack, so lets see if this works.
* Revert "[REVERT THIS] Add content to js to see if minify changes signing"
This reverts commit ecacc6830b.
* Revert "[REVERT THIS] Add a file A.ts to see if files are signed alphabetical or only non minified JS is signed"
This reverts commit 1ab6ea1873.
* Update name of incorrect task.
* Add failure handling for file integrity check
This is now what is hitting the EPERM and ENOENT errors.
We need to skip the check and see what happens from here, such as allowing to elevate via windows.
This also adds specific handlers for when we fail to download the SDK.
* elevate on windows
* Retry if no permission is available the first time
Need to double check that cancelling is handled properly now that its a promise reject
* respond to linter
* allow elevation on mac
* fix promise chain a bit, may need to await still
* Fix offline detection
8.8.8.8 blocks our requests. We can use microsoft.com instead, if our app is checking using microsoft, well, if microsoft is down we've learned that the app will have problems anyways , so if that is down it doesnt matter as much.
* fix variable name
* respond to linter
* respond to PR feedback, fix whitespace and ??
* add some fixes for lock, amend
* Use a separate lock for install vs installing
This will yield performance improvements.
* remove change that snuck in from another pr
* fix linter
* make functions not async if they dont need to be
* Hot fix for installation record tracking
We renamed it from 'key' to 'id' because the string 'key' is filtered out of all log events via vscode because 'key' looks like a secret. We did this in a rush and vendor testing did not have pre-existing machine state to realize the old data structure would still be in the local memory for pre-existing users. This converts the data in memory to the new type.
* Prevent adding corrupted installation records into the data
* Try to find the .NET SDK PATH on Mac if it Installs to Unexpected Location
Our highest failure for .NET SDK Installs is that the `InstallationValidator` which checks the hard-coded expected path of Installation does not find .NET in the expected location. Unfortunately, nobody has reported this.
To get to this point in the code of `Validation`, everything has to have succeeded, including the installer only returning and exiting 0, which it would only do upon installation success.
I'm guessing there is some undocumented case where the .NET SDK is installed to a different location, but I couldn't find where.
Here is our path check. https://github.com/dotnet/vscode-dotnet-runtime/blob/main/vscode-dotnet-runtime-library/src/Acquisition/WinMacGlobalInstaller.ts#L239
According to the installer source code and our documentation, it should only install to this location or with x64 during emulation.
```
<PkgInstallDirectory>/usr/local/share/dotnet</PkgInstallDirectory>
<x64EmulationPkgInstallDirectory>/usr/local/share/dotnet/x64</x64EmulationPkgInstallDirectory>
```
The 64 directory however, is only for when you install a 64-bit SDK on an ARM Mac. We wouldn't expect this to happen because we only install the SDK that matches os.platform(), so we should never install an x64 SDK on an ARM machine.
I'm curious to see where this path is and this will also help us track that.
It's a bit of a hack, but I think it will have the desired outcome.
* Remove extra check for global since this only happens in global code
* Change 'InstallKey' terminology to 'InstallId'
VS Code has a filter for the string 'key' -- this causes our telemetry to be dropped when referencing the term 'install key' as it thinks it may be a secret. 48ba01b589/src/vs/platform/telemetry/common/telemetryUtils.ts (L336)
Id is safe. I talked with the vscode team and there is no good alternative besides turning off all telemetry safety scanning, which I didn't want to do.
Hence, 'key' --> 'id'
* rename key to id
* rename remaining instances of 'id'
* Improve names of types
* Interpret Installer Exit Codes
These will give the user more actionable items, and also increase the chances the post a log so we can help them if something goes wrong and understand the issue better.
* Add page fault message
* Add Timeout error code
* Use Native Code Message
* Use err message from https://learn.microsoft.com/en-us/windows/win32/debug/system-error-codes--1300-1699-
* Use particular error code from windows
* Use the official error message.
* Add failure handling for file integrity check
This is now what is hitting the EPERM and ENOENT errors.
We need to skip the check and see what happens from here, such as allowing to elevate via windows.
This also adds specific handlers for when we fail to download the SDK.
* missing semicolon
* Make Command Executor return an object instead of string
This prevents us from having to run commands twice to get the status and stdout.
Caution must be taken to make sure that the assumptions here in what is returned as stdout vs stderr are correct.
Furthermore, when looking at the code, I saw something that looked just plain incorrect, so I will need to investigate that by adding the TODO before merge
* Fix Execution Type Parsing
* fix linting issue
* Fix line parsing logic
* Fix tests
* Run icacls to set execute permissions
* move to sha256
* remove unnecessary code
* Remove todo that should go in another PR
* Remove redundant trim
* rename class CommandProcessorOutput since it didnt match CommandExecutor
* Rename class file
* Report json object for failed permissions.
* Fix test to use correct folder
* fix lint
* Fix installer file deletion
* respond to linter
* Try to Add a Timeout for Installer.Exe
This will allow us to not count times when users never respond to the installer as a failure and or remediate when the installer is taking too long.
* Add .editorconfig for style settings
* Update Linux Apt Commands to Handle Lock Timeouts
We also pass the -y flag when it can help push through updates.
Apt will return 100 whenever there is any failure. aa56836331/apt-private/private-cmndline.cc (L602)
This means the errors we are seeing are not just due to locking errors. But, perhaps this will help.
Also this will try to run update twice, to see if that 2nd try will work. It seems a bit silly, but it does actually seem to improve the chance of success when something else is running on the system.
* Update distro failure message b to have the same information as the other failure message
* Fix package is available check
* Remove verbosity as it is not helpful or needed information to help customers
* Remove -y
it doesnt do anything for update lol
* fix json
* try forking
* just rely on node timeout
* update test data
* Handle .net installer timeouts
* remove unnecessary import
* Fix tests
* Include stderr for more failure detail
* fix options check
* Fix option
* Include options that are needed to preserve the working env on a general command execution
* Remove hash as it already gets anonymized by the regular expression
* Increase lock timeout time as now with waiting for lock it can take longer
* Add passive option to automate dialog
* Skip extra hashing thats handled by the regular expression
* Add a timeout to kill our subprocesses that hold onto the lock
* Timeout default is seconds not minutes
* Make code more clear to read by moving else after return into else
* Improve lock file handling on install tracker
* Adhere to style convention with release() block
* add back in -y to automate
---------
Co-authored-by: Ubuntu 1804 <you@example.com>
* Make Command Executor return an object instead of string
This prevents us from having to run commands twice to get the status and stdout.
Caution must be taken to make sure that the assumptions here in what is returned as stdout vs stderr are correct.
Furthermore, when looking at the code, I saw something that looked just plain incorrect, so I will need to investigate that by adding the TODO before merge
* Fix Execution Type Parsing
* fix linting issue
* Fix line parsing logic
* Fix tests
* Run icacls to set execute permissions
* move to sha256
* remove unnecessary code
* Remove todo that should go in another PR
* Remove redundant trim
* rename class CommandProcessorOutput since it didnt match CommandExecutor
* Rename class file
* Report json object for failed permissions.
* Fix test to use correct folder
* fix lint
* Fix installer file deletion
* respond to linter
* Try to Add a Timeout for Installer.Exe
This will allow us to not count times when users never respond to the installer as a failure and or remediate when the installer is taking too long.
* Add .editorconfig for style settings
* try forking
* just rely on node timeout
* Handle .net installer timeouts
* remove unnecessary import
* fix options check
* Fix option
* Include options that are needed to preserve the working env on a general command execution
* Add passive option to automate dialog
* dont use english url
Noah Gilson
andysleigh
Noah Gilson