EventPipe support for ProcessInfo #87562
Conversation
LakshanF
requested review from
elinor-fung
and removed request for
MichalStrehovsky
|
Tagging subscribers to this area: @tommcdon Issue DetailsMissing process information - work by @elinor-fung to support
|
|
Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas Issue DetailsMissing process information - work by @elinor-fung to support
|
| if (PalInterlockedCompareExchangePointer((void**)(&entrypoint_assembly_name), (void*)(entrypoint_assembly_name_local), nullptr) != nullptr) | ||
| delete[] entrypoint_assembly_name_local; | ||
| } | ||
| return reinterpret_cast<const char*>(entrypoint_assembly_name); |
There was a problem hiding this comment.
Is always returning the process name (minus extension) what we want? So if someone renames the executable or if someone uses a nativeaot library, they would get the name of the process which would not match the assembly.
There was a problem hiding this comment.
I think for the first iteration, using the process name without extension is fine. We can enhance if user feedback requests this.
| const wchar_t* process_name_const = wcsrchr(wszModuleFileName, DIRECTORY_SEPARATOR_CHAR); | ||
| if (process_name_const != NULL) { | ||
| process_name_const++; | ||
| } | ||
| size_t len = -1; | ||
| const wchar_t* extension = wcsrchr(process_name_const, '.'); | ||
| if (extension != NULL) { | ||
| len = extension - process_name_const; | ||
| } |
There was a problem hiding this comment.
You could define something like _tcslen / _tcsrchr based on the platform and then share this logic for finding the start of the name and its length.
There was a problem hiding this comment.
Added this to the tracking issue, #87069
ghost
locked as resolved and limited conversation to collaborators
Missing process information - work by @elinor-fung to support
commandLineand addedmanagedEntrypointAssemblyName. Enabled the existing diagnostic tests related to this.