UTF-8 vs current code page inconsistency

[PhilippeM1970]

Describe the bug
When I download a file using BlockBlobClient::DownloadTo, I need to convert CString to std::string using CP_UTF8 in CW2A call which is expected. However If want to upload a file using UploadFrom then the filename must not be in UTF-8 but in current code page (no page specified in CW2A.

Exception or Stack Trace
When I upload a file, I get an error during the call so convert back the filename to a wide filename if it contains accentuated characters.

To Reproduce
Call those functions with accentuated filename coming from CString object.

Code Snippet

void ReadBlob(const CString &fileName, const CString &blobName)
{
    const std::string blobNameUtf8(CW2A(blobName.GetString(), CP_UTF8));
    const std::string filename(CW2A(fileName.GetString(), CP_UTF8)); // *** Must have CP_UTF8 here

    BlockBlobClient blockBlob = m_ContainerClient->GetBlockBlobClient(blobNameUtf8);

   blockBlob.DownloadTo(filename);
}

void WriteBlob(const CString &fileName, const CString &blobName)
{
    const std::string blobNameUtf8(CW2A(blobName.GetString(), CP_UTF8));
    const std::string filename(CW2A(fileName.GetString())); // *** Must NOT have CP_UTF8 here

     BlockBlobClient blockBlob = m_ContainerClient->GetBlockBlobClient(blobNameUtf8);

     blockBlob.DeleteIfExists();
     blockBlob.UploadFrom(filename);
}

Expected behavior
Consistent use of UTF-8 in API.

Screenshots
If applicable, add screenshots to help explain your problem.

Setup (please complete the following information):

• OS: Windows 11
• IDE : Visual Studio 2022
• Version of the Library used 1.12.0 and 1.15.0

Additional context
With incorrect encoding, operation on file fails when filename contains accentuated characters.

Information Checklist
Kindly make sure that you have added all the following information above and checkoff the required fields otherwise we will treat the issuer as an incomplete report

• Bug Description Added
• Repro Steps Added
• Setup information Added

[Jinming-Hu]

In BlockBlobClient::UploadFrom(), filename is used in two places

1. used to construct Azure::Core::IO::FileBodyStream, where the filename is passed directly to Windows API CreateFile
2. used to construct FileReader, where file name is first converted from UTF-8 to UTF-16, then passed to open.

There doesn't seem to be a way to fix this without introducing a breaking change.

cc @RickWinter

[LarryOsterman]

FWIW, Behavior #2 is the "correct" behavior. The Azure SDK interprets input strings as being encoded as UTF-8, so behavior #1 above (FileBodyStream) is almost always incorrect (there are corner cases where it's acceptable, but they are rare).

Technically you're right, it could be considered a breaking change, but realistically case #1 never correctly worked. Imho, it's not a breaking change if you're going from "It never correctly worked" to "it now works correctly", even if there's a semantic change.

[Jinming-Hu]

@LarryOsterman is your team going to fix it?

[Jinming-Hu]

as confirmed by @LarryOsterman , this is an issue of the implementation of Azure::Core::IO::FileBodyStream, adding tag Azure.Core