[cdt-dev] CDT breakpoints and debug adapters throw NullPointerException

[Stephen Williams <shwilliams54321@xxxxxxxxx>]

Hi,

I posted this message to the CDT forum yesterday, but I realized this may be a
more appropriate list as this is a query about what I believe to be a CDT bug.

I've created an EFS implementation so my projects can work with remote resources.

This is going quite nicely, but I ran into a problem with remote debugging with
ResumeAtLineAdapter and other debug-related objects.

When a debug session starts, my source look-up object returns an IFile whose

path contains a linked resource. CDT then nicely loads the remote source file
into an editor. Problems occur (NullPointerException), though, when the

ResumeAtLineAdpater is getting set up (and similar issues happen when I try
to set breakpoints).

ResumeAtLineAdapter calls CDebugUIUtils.getEditorFilePath() to get the path

associated with an IEditorInput object like this:

    public static String getEditorFilePath( IEditorInput input ) throws CoreException {

      if ( input instanceof IFileEditorInput ) {
  -->   return ((IFileEditorInput)input).getFile().getLocation().toOSString();

      }

That call results in a call to getLocation() (see above), which in turn calls:

  FileSystemResourceManager.locationFor(IResource)


which in turn calls this method:

  FileStoreRoot.localLocation(IPath, IResource)

The problem is, FileStoreRoot will always return null for a linked resource, which will result
in an unhandled NullPointerException:

  /**
   * If this root represents a resource in the local file system, this path

   * represents the root location.  This value is null if the root represents
   * a non-local file system

   */
  private IPath localRoot = null;

    .
    .
    .
  IPath localLocation(IPath workspacePath, IResource resource) {
    if (localRoot == null)

      return null;

It seems to me that this is a bug in getEditorFilePath(). It should not be

calling getLocation(), as that assumes that the resource is local, which,
in this case, it isn't.

If anyone has a suggestion for a workaround, I would greatly appreciate it.

I could open a bug on this if it seems warranted. I think I did see a related

bug about not being able to relocate a workspace due to breakpoints using
absolute paths (or something similar).

Thanks!


--Steve