Introduce SafeProcessHandle.Start and ProcessId by adamsitnik · Pull Request #126192 · dotnet/runtime

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik · 2026-03-27T09:47:12Z

This is one of very few PRs that I recently wrote myself without using AI. The reason is that I've failed to write the right prompt and TBH and I was also afraid of some subtle hallucinations being introduced.

I've just moved all the helpers from Process.[Unix/Win32/Windows].cs into ProcessUtils.$platform.cs and moved the logic responsible for launching processes with given handles to SafeProcessHandle.StartCore.

First two commits are more or less cut and paste.

The third commit is most important: 0028e06 as it contains some non-obvious changes (see my other comments).

The fourth commit is just exposing ProcessId.

The fifth commit are some basic tests.

The API was approved in #125838 (comment).

This PR will allow me to implement the new "scenario" methods without using Process directly and referencing a tone of code. Example: #126078

Once this is merged, I am going to send other PR:

SafeProcessHandle.Kill and Signal
SafeProcessHandle.WaitForExit*

fixes#126107

… (all tests passing)

dotnet-policy-service · 2026-03-27T09:48:19Z

Tagging subscribers to this area: @dotnet/area-system-diagnostics-process
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot reviewed 22 out of 22 changed files in this pull request and generated no comments.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated no new comments.

adamsitnik · 2026-03-27T15:10:08Z

@copilot invoke the code-review skill and post the analysis/results as a comment on this PR

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 5 comments.

The ProcessWaitState.Holder was created outside the read lock in the refactored ForkAndExecProcess, introducing a race condition: 1. Fork+exec child (inside read lock) 2. Release read lock 3. SIGCHLD fires → OnSigChild takes write lock → CheckChildren runs 4. Child is NOT yet in s_childProcessWaitStates → SIGCHLD is consumed without reaping the child 5. ProcessWaitState.Holder created (too late) 6. WaitForExit waits on _exitedEvent that will never be set → HANG The fix mirrors the original Process.Unix.cs code: create the ProcessWaitState.Holder while still holding the read lock, so the child is in s_childProcessWaitStates before SIGCHLD can be processed. Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/325db55f-9e1f-43bd-abec-1b60c0b2b5ce Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

tmds · 2026-03-30T07:23:41Z

@jkotas @adamsitnik I wonder if you have some insights why Process ends up being difficult to trim?

adamsitnik · 2026-03-30T08:20:48Z

@rokonec @eiriktsarpalis Could you please review the Windows part? In the first commit (968d39a) I've just moved the code to other files to allow for exposing new APIs that don't have a dependency on Process. I've added GH comments in places where I've introduced changes to Windows implementation.

adamsitnik · 2026-03-30T08:44:31Z

I've not investigated it myself (yet).

rokonec

LGTM

eiriktsarpalis

A few minor post-merge comments:

eiriktsarpalis · 2026-03-30T12:00:15Z

+
+ if (!startInfo.UseShellExecute)
+ {
+ if (childInputHandle is null && !OperatingSystem.IsAndroid())


Out of curiosity, what's specific about android in this check?

Android does not support Console.OpenStandard[Input/Output/Errror]Handle methods as it does not support Console, but it still has some standard input/output/error handles, so we just pass null here and the native layer takes care of it.

eiriktsarpalis · 2026-03-30T12:01:11Z

 return true;
 }
+
+ // On Unix, we don't use process descriptors yet, so we can't get PID.


What does "yet" mean in this context? Is it something we haven't implemented in the PAL? Do we need to link to an open issue?

I intend to extend SafeProcessHandle with process descriptors support for 11, but just not in this PR.

eiriktsarpalis · 2026-03-30T12:03:28Z

 public sealed partial class SafeProcessHandle : SafeHandleZeroOrMinusOneIsInvalid
 {
 internal static readonly SafeProcessHandle InvalidHandle = new SafeProcessHandle();
+ private int _processId = -1;


Seems like a prime use case for using the new field keyword.

eiriktsarpalis · 2026-03-30T12:26:19Z

+ : new("sh") { ArgumentList = { "-c", "exit 42" } };
+
+ using SafeProcessHandle processHandle = SafeProcessHandle.Start(startInfo);
+ Assert.NotEqual(0, processHandle.ProcessId);


Should there be a test checking for the validation code in this property getter?

…ndForget Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6ec5336e-ce95-4d76-ad71-62ab2f1b4a81 Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com>

…n StartAndForget" This reverts commit c6a0adb.

jkotas · 2026-03-30T21:37:24Z

It always drags in a lot of performance counter related functionality and the dependencies are not straightforward to break without observable behavior changes. (It may be possible to improve this, it is just not straightforward.)

For example, here is a dependency chain from https://github.com/MichalStrehovsky/sizoscope for a simple app that just does Process.Start("notepad.exe").WaitForExit()

jkotas · 2026-03-30T21:38:23Z

Opened #126332

…te) (#126314) Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: adamsitnik <6011991+adamsitnik@users.noreply.github.com> Co-authored-by: jkotas <6668460+jkotas@users.noreply.github.com> Co-authored-by: Jan Kotas <jkotas@microsoft.com>

adamsitnik added 2 commits March 26, 2026 21:18

refactor the code and introduce SafeProcessHandle.Start: Windows part…
968d39a
… (all tests passing)

refactor the code and introduce SafeProcessHandle.Start: Unix part
d12149f

adamsitnik added the NO-REVIEW Experimental/testing PR, do NOT review it label Mar 27, 2026

CopilotAI review requested due to automatic review settings March 27, 2026 09:47

github-actionsBot added the area-System.Diagnostics.Process label Mar 27, 2026

dotnet-policy-serviceBot assigned adamsitnik Mar 27, 2026

Copilotstarted reviewing on behalf of adamsitnik March 27, 2026 09:49 View session

CopilotAI reviewed Mar 27, 2026
View reviewed changes

adamsitnik force-pushed the sph-start branch 2 times, most recently from 2d1b324 to c2eda1dCompare March 27, 2026 14:39

CopilotAI review requested due to automatic review settings March 27, 2026 14:39

Copilotstarted reviewing on behalf of adamsitnik March 27, 2026 14:40 View session

CopilotAI reviewed Mar 27, 2026
View reviewed changes

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment thread...ries/System.Diagnostics.Process/src/Microsoft/Win32/SafeHandles/SafeProcessHandle.Windows.cs

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment thread...braries/System.Diagnostics.Process/src/Microsoft/Win32/SafeHandles/SafeProcessHandle.Unix.cs Outdated

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment threadsrc/libraries/System.Diagnostics.Process/src/System/Diagnostics/Process.FreeBSD.cs

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment threadsrc/libraries/System.Diagnostics.Process/src/System/Diagnostics/Process.Unix.cs Outdated

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment threadsrc/libraries/System.Diagnostics.Process/src/System/Diagnostics/ProcessUtils.Unix.cs

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment threadsrc/libraries/System.Diagnostics.Process/src/System/Diagnostics/ProcessUtils.Windows.cs

adamsitnik commented Mar 27, 2026
View reviewed changes

Comment threadsrc/libraries/System.Diagnostics.Process/src/System/Diagnostics/ShellExecuteHelper.cs

adamsitnik removed the NO-REVIEW Experimental/testing PR, do NOT review it label Mar 27, 2026

adamsitnik changed the title ~~Introduce SafeProcessHandle.Start~~Introduce SafeProcessHandle.Start and ProcessIdMar 27, 2026

adamsitnik added this to the 11.0.0 milestone Mar 27, 2026

adamsitnik requested a review from Copilot March 27, 2026 15:16

Copilotstarted reviewing on behalf of adamsitnik March 27, 2026 15:17 View session

CopilotAI reviewed Mar 27, 2026
View reviewed changes

adamsitnik added 2 commits March 27, 2026 18:00

the tricky part
0028e06

implement ProcessId
c11b87c

adamsitnik requested a review from tmds March 29, 2026 11:26

build-analysisBot mentioned this pull request Mar 29, 2026
Unable to pull image from mcr.microsoft.com #117164
Closed

This was referenced Mar 30, 2026
Implement SafeProcessHandle.Kill and Signal #126291
Closed
Implement SafeProcessHandle.WaitForExit* methods #126293
Closed

tmds reviewed Mar 30, 2026
View reviewed changes

Comment thread...braries/System.Diagnostics.Process/src/Microsoft/Win32/SafeHandles/SafeProcessHandle.Unix.cs

tmds reviewed Mar 30, 2026
View reviewed changes

Comment thread...braries/System.Diagnostics.Process/src/Microsoft/Win32/SafeHandles/SafeProcessHandle.Unix.cs

tmds reviewed Mar 30, 2026
View reviewed changes

Comment thread...braries/System.Diagnostics.Process/src/Microsoft/Win32/SafeHandles/SafeProcessHandle.Unix.cs

tmds approved these changes Mar 30, 2026
View reviewed changes

adamsitnik requested review from eiriktsarpalis and rokonec March 30, 2026 08:17

rokonec approved these changes Mar 30, 2026
View reviewed changes

adamsitnik merged commit dee9005 into dotnet:mainMar 30, 2026
93 of 97 checks passed

CopilotAI mentioned this pull request Mar 30, 2026
Implement SafeProcessHandle.Kill and Signal #126313
Merged

eiriktsarpalis reviewed Mar 30, 2026
View reviewed changes

CopilotAI mentioned this pull request Mar 30, 2026
Follow-up to #126192: address review comments (doc nit, ShellExecuteHelper refactor, field keyword, test, trimming, waitStateHolder fix) #126314
Merged

github-actionsBot mentioned this pull request Mar 30, 2026
Move polyfills for !NET targets to src/libraries/Common/src/Polyfills/ #126287
Merged

adamsitnik added a commit that referenced this pull request Mar 30, 2026
Revert "Sync with main (PR#126192) and use SafeProcessHandle.Start i…
0f411da
…n StartAndForget" This reverts commit c6a0adb.

jkotas mentioned this pull request Mar 30, 2026
Investigate improving trimability of System.Diagnostics.Process #126332
Closed

CopilotAI mentioned this pull request Mar 31, 2026
Improve trimability of System.Diagnostics.Process by lazily initializing remote machine support #126338
Merged

dotnet-maestroBot mentioned this pull request Mar 31, 2026
[main] Source code updates from dotnet/runtime dotnet/dotnet#5799
Merged

CopilotAI mentioned this pull request Apr 5, 2026
[WIP] Implement SafeProcessHandle.WaitForExit methods #126563
Closed
11 tasks

github-actionsBot locked and limited conversation to collaborators Apr 30, 2026