Redundant double debug registrations for plugins

[colin-grant-work]

Bug Description:

We register debuggers once in the contribution handler and once in the plugin context.

In the PluginDebugService, the first registration puts the contribution in the debuggers array and the second registration puts the contribution in the contributors map so that we end up with completely parallel lists, vz. the output of a log in the init method of the PluginDebugService that sets a timeout to log { debuggers: this.debuggers.map(contrib => contrib.type), contributors: Array.from(this.contributors.keys()) }:

{
  debuggers: [
    'extensionHost',
    'node2',
    'pwa-node',
    'node-terminal',
    'pwa-extensionHost',
    'pwa-chrome',
    'pwa-msedge',
    'node'
  ],
  contributors: [
    'extensionHost',
    'node2',
    'pwa-node',
    'node-terminal',
    'pwa-extensionHost',
    'pwa-chrome',
    'pwa-msedge',
    'node'
  ]
}

Additional Information

• Theia Version: 41a2f5b

[alvsan09]

@colin-grant-work, I am not quite able to follow the last part,
Could you please point me to the logging of the identical entries under the PluginDebugService?

The types for the contributors vs debuggers is different although I agree we should try to merge them if possible.

[colin-grant-work]

@alvsan09, there isn't a log like the one I displayed in the current code. To generate it, I added this line to the init method:

setTimeout(() => console.log({ debuggers: this.debuggers.map(contrib => contrib.type), contributors: Array.from(this.contributors.keys()) }), 10_000);

They are indeed different types, but the condition for the two registrations is the same: the presence of a .contributes.debuggers field, so the registrations from the plugin context will always be a (not necessarily proper) subset of the registrations from the contribution handler. Perhaps the map is the correct approach, and then we can replace the registrations from the contribution handler (which are just data with no code), with the registrations from the plugin context (which have the same data, but also executable code attached) as the latter come in?

[colin-grant-work]

Fixed by #10910.