Ugh. I was worried about this. This is from this CL:

https://chromium-review.googlesource.com/#/c/394827/

sadrul@ - it doesn't look like *_manifest.json always exists, we need to make it optional.

 Issue 656694  has been merged into this issue.

Ugh, my bad, sorry :(

+sky@

From what I can see, the gn flag 'enable_package_mash_services' is always turned on for chormeos [1]. When this flag is turned on, the 'chrome/app:service_manifests' target is added as a data-dependency to chrome [2]. And that target brings in all the targets that generate the manifests files [3]. So as long as chrome is being built, I don't see how the manifests files are not generated.

Does the bot turn off the enable_package_mash_services gn flag?

[1] https://cs.chromium.org/chromium/src/chrome/common/features.gni?q=chrome/common/features.gni&sq=package:chromium&dr&l=47).
[2] https://cs.chromium.org/chromium/src/chrome/BUILD.gn?q=chrome/BUILD.gn&sq=package:chromium&dr&l=253
[3] https://cs.chromium.org/chromium/src/chrome/app/BUILD.gn?q=chrome/app/BUILD.gn&sq=package:chromium&dr&l=395

The second link is inside the chrome_inital target:

executable("chrome_initial") {

Which possibly isn't getting built in SimpleChromeWorkflow?

The third link is in chrome/app/BUILD.gn which is very likely not getting built.

From what I can see, 'chrome_initial' is a dependency of 'chrome' target: https://cs.chromium.org/chromium/src/chrome/BUILD.gn?q=chrome/BUILD.gn&sq=package:chromium&dr&l=97 (which also in fact directly adds a dependency on service_manifests) .. so I have no idea why the manifest files aren't being generated.

Ignore that last comment, I dind't look closely enough. "chrome_initial" appears to be a dep for group("chrome"), so I would expect it to be built.

I'm going to try a clean local build now.

So, we don't seem to be generating any /*_manifest.json files for a typical SimpleChrome build. i.e.:

(sdk samus 8890.0.0) $ ninja -j 1000 -l 10 -C out_$SDK_BOARD/Release chrome chrome_sandbox nacl_helper
(sdk samus 8890.0.0) $ find out_$SDK_BOARD/Release -name *_manifest.json
<no output>

sadrul@ - We need to either make this optional or revert it asap then figure out why the expected files are not getting generated.

So, all 4 of the manifests listed in group("service_manifests") have the pattern:

source = "${root_out_dir}/Packages/content_browser/manifest.json"

i.e. the generated files do not follow the pattern "*_manifest.json".

I am going to go ahead and revert this so that we can unbreak the PFQ asap.

OK, thanks for taking care of it. I will upload with a better fix. Sorry about the trouble.

So, this landed and should fix the $subject bug?

https://chromium-review.googlesource.com/#/c/399566/

In the future, please use BUG= so the ticket gets autoupdated.

Should we expect the next runs to succeed? (I see a current round of PFQ builders that looks doomed to fail still, but after that?)

Yes. The next PFQ run should at least not fail due to this bug. We should
start a new run as soon as this one completes (and we confirm that we only
had SimpleChrome failures or flake).

I don't know an easy way to change the commit message in gerrit (without
downloading the generated revert patch and re-uploading)?

There's an 'Edit' button near the lower half of the page, alongside 'Open All' and 'Diff against: <drop down>'.

Oh, huh, and then it looks like I can click on 'Commit Message' and edit
it. Obscure, but handy, thanks!

Builds are starting to look green.