Set callconv in unicodedata callback and update doc

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks

[am11]

toupper/lower in C++ only transforms the case of ASCII characters. It's different than what we expect in coreclr-pal and generally in C#. This change switches to minpal's toupper/lower which takes care of chars with opposite case in 0..0xFFFF range.

Repurposed: set callconv for x86 and update docs.

[dotnet-policy-service]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Is this something we want? The functions in corehost are used for parsing arguments, environment variables, path and runtimeconfig. Does any of these has normalization of non-ASCII text?

[am11]

I had this on my radar for a while #44466 (comment). Now that we have minipal, we can use shared implementation without much of a hassle.

[vitek-karas]

@am11 could you please measure the impact of this change to the size of the apphost executable? We don't really care about size of dotnet executable or hostfxr,hostpolicy, but apphost is part of every app, so we want it as small as possible.

[am11]

It increases the size of apphost by 16.6 KB on osx-arm64 (before: 125328, after: 141936 in bytes).

If this is unacceptable. I can drop the corehost changes from the PR and keep the others (doc update and win32 build fixes of unicodedata).

[jkotas]

Does this fix a bug in any observable behavior?

As far as I can tell, all places that use to_lower should work just fine with the C ASCII-only behavior. I do not think we want to take the binary size hit unless it fixes something material.

[am11]

@jkotas in the linked issue, it was found when runtime was cloned at non-ASCII path and host was trying to compare lowered hostfxr path with the lowered path returned on managed side in unit tests. Other than that, nethost uses it during the arg parsing (which we should anyway avoid). I’ll drop it and repurpose the PR.

[jkotas]

Thanks