Many thanks for the report!

It seems like a bug in our recipe to rerun the test.
a708b081adfe was skipped and not tested.

https://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_variable_10.11/builds/2727

sky@, dpranke@: Can you help confirm whether below is the expected behavior for "mb.py analyze"? If not, should we add a new exception filter?

The detail is in nested in the wrapping step "test a708b08" in https://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_variable_10.11/builds/2727


analyze input:
{
  "additional_compile_targets": [], 
  "files": [
    "content/browser/resources/media/client_renderer.js", 
    "content/browser/resources/media/manager.js", 
    "content/browser/resources/media/media_internals.css", 
    "content/browser/resources/media/media_internals.html"
  ], 
  "test_targets": [
    "content_browsertests"
  ]
}

analyze output:
{
  "compile_targets": [], 
  "status": "No dependency", 
  "test_targets": []
}

No, I wouldn't necessarily expect that result. You might be running into  bug 639328 , but I need to look further to see for sure and/or figure out if something else is going on.

I think the problem here is that the dependencies are under-specified. I'm not entirely sure whether the files aren't referenced at all at that revision, or whether they're referenced in a way we're not smart enough to compute. The fix would be to make sure they are explicitly listed as dependencies in a GN file.

stgao@, I'm going to bounce this back to you to decide where to go from here.

(many thanks Dirk for the investigation!)

Assign to tommi@ -- owner of the directory.
tommi@, could you help out fix the problem here?

As Dirk mentioned, the four files below have to be listed as dependencies in a GN file:
    "content/browser/resources/media/client_renderer.js", 
    "content/browser/resources/media/manager.js", 
    "content/browser/resources/media/media_internals.css", 
    "content/browser/resources/media/media_internals.html"

Otherwise, any change to them will not run content_browsertests in CQ and potentially causes a breakage on Waterfall like https://chromium-review.googlesource.com/c/513806/

Henrik - can you perhaps help with this?

I'm afraid no, I don't have any expertise in how analyze is implemented. I agree it's important to get content_browsertests to run in this case. I have been bitten by similar errors in the past (analyze would benefit from better test coverage).

Assigning to brettw@ since he did a fix for grit in the bug mentioned in #3 ( bug 639328 ).

@tommi, @kjellander - this isn't a complicated fix, or even a code fix, you just need to add the files listed in comment #5 as inputs to the //content:resources target.

hbos - can you take a look? (hopefully a simple thing to fix)