1910497 - Debugger.Source.text is reporting "[no source]" for worker modules loaded via ChromeUtils helpers

Status: RESOLVED FIXED

(Core :: JavaScript Engine, task, P3)

Description

[Alexandre Poirot [:ochameau]]

Debugger's Source text attribute is returning "[no source]" for worker modules loaded via ChromeUtils.importESModule/defineESModuleGetters, at least when using { global: "current" } argument.
Note that it works fine when loading the modules via static imports.

I'll provide a mochitest.

Comment 1

[Alexandre Poirot [:ochameau]]

Attachment: Bug 1910497 - Test missing source text content on worker modules loaded via ChromeUtils helpers.

Comment 2

[Matthew Gaudet (he/him) [:mgaudet]]

Perhaps something to put on your backlog?

Comment 3

[Nicolas B. Pierron [:nbp — off until 29-09]]

I recall something like that being made on purpose, as we expected that none of Firefox functions should rely on the source, with no consideration for debugging. However I would have expected it to be the same independently of workers.

Arai, Yulia, Ted, any insight into what might be happening here?

Comment 4

[Tooru Fujisawa [:arai]]

(In reply to Nicolas B. Pierron [:nbp] from comment #3)

I recall something like that being made on purpose, as we expected that none of Firefox functions should rely on the source, with no consideration for debugging. However I would have expected it to be the same independently of workers.

For system principal, we had been discarding the source, but it doesn't apply here.

What happening here is that the the source is marked "retrievable", but the runtime doesn't have the source hook:

https://searchfox.org/mozilla-central/rev/e637cd67f98ed4272d13a96db9a7674689122dcb/js/src/vm/JSScript.cpp#836,857-860

class ScriptSource::LoadSourceMatcher {
...
  template <typename Unit>
  bool operator()(const Retrievable<Unit>&) {
    if (!cx_->runtime()->sourceHook.ref()) {
      *loaded_ = false;

This is caused by setSourceIsLazy, which expects the runtime to have the XPCJSSourceHook, but the worker runtime doesn't have it.

https://searchfox.org/mozilla-central/rev/e637cd67f98ed4272d13a96db9a7674689122dcb/js/xpconnect/loader/mozJSModuleLoader.cpp#865-867,876-877

nsresult mozJSModuleLoader::LoadSingleModuleScriptOnWorker(
    SyncModuleLoader* aModuleLoader, JSContext* aCx,
    JS::loader::ModuleLoadRequest* aRequest, MutableHandleScript aScriptOut) {
...
  CompileOptions options(aCx);
  ScriptPreloader::FillCompileOptionsForCachedStencil(options);

https://searchfox.org/mozilla-central/rev/e637cd67f98ed4272d13a96db9a7674689122dcb/js/xpconnect/loader/ScriptPreloader.cpp#920-921,930

void ScriptPreloader::FillCompileOptionsForCachedStencil(
    JS::CompileOptions& options) {
...
  options.setSourceIsLazy(true);

https://searchfox.org/mozilla-central/rev/e637cd67f98ed4272d13a96db9a7674689122dcb/js/xpconnect/src/XPCJSRuntime.cpp#3005-3006

mozilla::UniquePtr<XPCJSSourceHook> hook(new XPCJSSourceHook);
js::SetSourceHook(cx, std::move(hook));

So the solution would be to skip the setSourceIsLazy for worker case.

Comment 5

[Tooru Fujisawa [:arai]]

Attachment: Bug 1910497 - Do not use lazy source in ChromeUtils.importESModule in worker. r?bthrall!

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/b6440b259f94
Do not use lazy source in ChromeUtils.importESModule in worker. r=bthrall

Comment 7

[pstanciu]

https://hg.mozilla.org/mozilla-central/rev/b6440b259f94