It looks like we have a timing issue with how modules are loaded, which is affecting how the sources appear in the debugger. STR: 1. go to https://cdn.rawgit.com/jakearchibald/a298d5af601982c338186cd355e624a8/raw/aaa2cbee9a5810d14b01ae965e52ecb9b2965a44/ 2. add a breakpoint in (index) on line 6 addTextToBody 3. refresh 4. step into addTextToBody and pause in utils.js 5. add a breakpoint in utils.js#3 6. remove the breakpoint in (index) 7. refresh Expected: pause in utils.js Actual: don't pause, and utils.js is not in the source tree

Jason, if this is something you think we ought to just knock out, let me know in the weekly meeting. It may require some pair debugging to pin it down.

Sounds good. I believe Ted had some ideas of where to get started here as well.

My suspicion is this may be related to Bug 1440269 where we added a mechanism to defer binding of the <script> element to the JSScript. This info is later used by the debugger's matching logic. See things hanging off |hideScriptFromDebugger|.

This is almost certainly my fault. I'll try and get to this tomorrow.

The problem was that the process of associating source elements with a script and exposing the script to the debugger only happened for the root module in a module graph. The patch does this recursively for a module and all of its imports (module requests form a tree even when the graph has cycles). I tested that this fixed the problem locally.

Comment on attachment 9003119 [details] [diff] [review] bug1469599-associate-module-graph Review of attachment 9003119 [details] [diff] [review]: ----------------------------------------------------------------- r+ ::: dom/script/ScriptLoader.cpp @@ +967,5 @@ > nsresult > +ScriptLoader::AssociateSourceElementsForModuleTree(JSContext* aCx, > + ModuleLoadRequest* aRequest) > +{ > + // Preloaading can cause JS scripts to be compiled before DOM script element Typo: "Preloaading"

Pushed by jcoppeard@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/3f1d6721cab0 Associate all module scripts in a module graph with DOM elements before execution r=hsivonen

It looks like the STR is still failing, but that likely means there are multiple underlying issues here.

re-opening as i can still reproduce the issue.

never mind, i was testing with artifact builds and forgot to wait for the patch to land on central. Thanks for fix! https://user-images.githubusercontent.com/254562/44485540-e8d06000-a61e-11e8-843b-ae35f269040d.png