Disable new warnings introduced with clang 18 by omajid · Pull Request #99811 · dotnet/runtime

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

omajid · 2024-03-15T12:12:21Z

clang 18 introduces -Wswitch-default, which requires that every switch must have a default branch. We can add missing default in switches, but the other option -Wcovered-switch-default complains if all the cases in a switch are exhaustive and default doesn't do anything. So disable one of these mutually exclusive warnings.

We will also need to merge in the changes from
dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

clang 18 introduces `-Wswitch-default`, which requires that every switch must have a `default` branch. We can add missing `default` in switches, but the other option `-Wcovered-switch-default` complains if all the cases in a switch are exhaustive and `default` doesn't do anything. So disable one of these mutually exclusive warnings. We will also need to merge in the changes from dotnet/arcade#14572 to actually try and use clang-18/clang++-18.

am11 · 2024-03-15T15:09:00Z

How feasible it is to fix the warning instead of suppressing it? We will soon have CI legs with clang18 (dotnet/dotnet-buildtools-prereqs-docker#976).

omajid · 2024-03-15T15:24:14Z

Like I wrote in the commit message, it's an issue because 2 conflicting warnings are now both enabled with clang 18. It's not possible to modify the code to make both warnings go away. See my issue that I reported to clang folks: llvm/llvm-project#85233

We can pick which warning to disable (or turn off -Weverything). I picked disabling switch-default because it seems less useful of the two.

am11 · 2024-03-15T15:29:29Z

Ah, I missed that. Good investigation and I agree with your choice. 👍

MichalPetryka · 2024-03-15T15:31:37Z

I picked disabling switch-default because it seems less useful of the two.

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

omajid · 2024-03-15T15:39:42Z

Wouldn't it make more sense to keep that one and add default with unreachable asserts to enum switches?

I am leaning towards no

Suppose we have an enum + a switch that goes over it. Without using default, we can be sure that our switch is exhaustive and get compile-time detection if we miss a case. If we add default everywhere (what -Wswitch-default requires) + -Wno-covered-switch-default, we can add new enum entries and forget to handle it in code. That turns a compile-time check into a runtime assert, delaying the time until the someone working on a piece of code gets feedback on an issue.

That said, I don't own any code in this repo, so I am happy to do what the maintainers tell me they want 😄

am11 · 2024-03-15T15:40:24Z

covered-switch-default is an old diagnostic which the code conforms to and other compilers also care for. switch-default is a new clang18 thing, and it is mutually exclusive. So that's why it is less useful.

janvorli · 2024-03-15T15:57:28Z

I agree with @omajid here. Let's keep this change as is.

janvorli

LGTM, thank you!

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - dotnet#94782 - dotnet#99811

This is a targeted backport from a few other PRs that makes it possible to build dotnet/runtme's 8.0 branch on Fedora 40 which includes clang 18. - dotnet/arcade#14572 - #94782 - #99811

ghost added the needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners label Mar 15, 2024

dotnet-policy-serviceBot added the community-contribution Indicates that the PR has been added by a community member label Mar 15, 2024

omajid force-pushed the clang-18 branch from 4b1c81d to db35f14Compare March 15, 2024 12:45

am11 requested a review from janvorli March 15, 2024 15:29

am11 added area-Infrastructure-coreclr Only use for closed issues and removed needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners labels Mar 15, 2024

janvorli approved these changes Mar 15, 2024
View reviewed changes

janvorli merged commit ac1b478 into dotnet:mainMar 15, 2024

omajid mentioned this pull request Mar 25, 2024
[release/8.0-staging] Support building against clang 18 #100258
Merged
4 tasks

github-actionsBot locked and limited conversation to collaborators Apr 15, 2024