Intellisense hang - Reliable reproduction

[pmolodo]

Environment

• OS and Version: Windows 11 23H2
• VS Code Version: 1.90.2
• C/C++ Extension Version: 1.20.5
• If using SSH remote, specify OS of remote machine: N/A

Bug Summary and Steps to Reproduce

Intellisense hang - Reliable reproduction

• OS and Version: Windows 11 23H2
• VS Code Version: 1.90.2
• C/C++ Extension Version: 1.20.5
• If using SSH remote, specify OS of remote machine: N/A

Bug Summary:
Infinite hang on "IntelliSense: Updating", and inability to browse source files

Steps to reproduce:

1. git clone https://github.com/pmolodo/vscode_cpp_intellisense_fail.git 
   cd vscode_cpp_intellisense_fail

   ...or just create an empty directory, manually download repro.ps1
   and [settings.json(https://github.com/pmolodo/vscode_cpp_intellisense_fail/blob/main/settings.json)] from the above repo into it, and cd in

2. powershell -ExecutionPolicy ByPass ./repro.ps1

   ...or inspect repro.ps1 and manually execute the commands, it's not long
3. This should launch a fresh vscode workspace for OpenUSD, with only the ms-vscode.cpptools extension installed
4. It should have already opened pxr\imaging\plugin\hdEmbree\renderer.cpp - if not, browse and open it
5. Wait for the file parser to finish
6. In the C++ output log, you should see that it's able to find most of the relevant source in the repo - at
   least enough to allow basic source browsing
7. IntelliSense, however, will keep spinning, stuck on "IntelliSense: Updating"
8. Try ctrl-clicking to browse to one of the headers - ie, pxr/imaging/plugin/hdEmbree/renderBuffer.h
9. Witness the nothing!

Expected behavior:

1. IntelliSense should finish, and at least basic file browsing should be possible

Configuration and Logs

• There's no c_cpp_properties.json in my example - it's using a fresh workspace with default settings. However, the default includePath=${workspaceFolder}\** should be enough for it to do basic discovery and file browsing, at least.
  Running C/C++: Log Diagnostics didn't have any obvious effect...?
• The output setting the language server logging to "Debug" are also in the above repo:
  cpptools.debug.log

Other Extensions

No response

Additional context

NOTE: the above reproduction was trying to browse a codebase that has not been built yet, and without all the dependencies, for easier / faster reproduction. However, the issue still happens with a complete / functioning build, with all dependencies. To reproduce this scenario (be prepared to wait!) on a "built" repo, with all dependencies, you can follow these steps:

• Install python 3.10, add it to your path
• Open a Visual Studio Developer Command Prompt
• cd into the OpenUSD directory
  NOTE: Make sure your OpenUSD directory absolute filepath isn't too long, or the build may
  hit the Windows max file path limit...
• pip install jinja2 pyopengl PySide6
• python build_scripts\build_usd.py _install -v

[Colengms]

Hi @pmolodo . I believe the issue here is that the project is complex and not properly configured for IntelliSense. Without sufficient configuration to successfully build the code, the IntelliSense parser appears to be spending an excessive amount of time trying to evaluate code it can't possibly evaluate and manage the abundance of errors that generates.

IntelliSense uses the EDG C/C++ front-end internally. In order to provide information on more than just the first few errors, it's being driven in such a way as to force it to continue parsing code despite encountering errors that would normally cause it to stop parsing.

I'd suggest first configuring the project by providing include paths and defines in c_cpp_properties.json. Or, since the project appears to be built with CMake, you could use CMake Tools (as a custom configuration provider to cpptools) to configure IntelliSense exactly as the code is built. Alternatively, if you are only looking for basic browsing support and don't require IntelliSense, you could set, "C_Cpp.intelliSenseEngine": "Tag Parser".

Obviously, IntelliSense should not stall indefinitely like that. We'd be open to addressing issues with specific code segments that cause the IntelliSense parser to misbehave (run slowly or fail to complete, etc.). But when not properly configured, a large project presents a huge volume of code to distill to identify issues from. Issues with complex codebases that only occur when not properly configured are not likely to get priority. So, I'm going to close this issue. Definitely let us know if you see incorrect behavior when a project is configured for IntelliSense.