Changing title - it seems to have nothing to do with the specific test. If I remove the test mention from the command and let it try to run all the SmokeTests  it still fails in the same way upon running the first test (instantly transitions to stopping the http server).

Dirk, can you help or help find another owner?

Not being able to run these tests means they rot and clank suffers. My end goal here is to:

1. Get the tests to work with GN (GYP will continue to break since AFAIK all other bots have moved away from it).
2. Make the android_blink_rel bot use GN

I can take a look and if this works for me (or not) when I'm back in the office in front of a linux machine w/ my devices.

However, I can't really own this. Mike, do you want to own this or find someone on blink infra to own it?

How different are the layout tests from chromium gtests?

The tests themselves are quite different.

I think the actual setup/teardown of the device for the tests is fairly similar, but is implemented in two different code bases (i.e., totally duplicated).

I think (but am not sure) that the problem Chris is seeing is a setup/teardown problem, so it would be good to get someone else to at least try to run the tests locally to see if they can repro.

Though, checking this now ... Chris's device is running J, but I think the builder is using K:

https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29/builds/47130

The tests are very sensitive to the device/os config. Chris, maybe try upgrading your phone if you can and see if that fixes things?

Ran with KitKat just now, observed same failing behavior. Log attached.

Deleted: out.txt 26.9 KB

out.txt 26.9 KB View Download

I want to comment that the GYP build is again regressed (this change [0] moves a bunch of files, updates .gn but not .gyp), but I'm puzzled by how the Nexus 4 bot seems to be green at the moment? 

The breaking change has position #396515. 
The last nexus 4 build [1] used change #397172. (I'm also synced here locally)


[0] https://codereview.chromium.org/1954363002
[1] https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29/builds/47875

When I attempt to gyp_chromium, I hit the error

find: `java/src/android/support/customtabs': No such file or directory
gyp: Call to 'find java/src/android/support/customtabs -name "*.java" | sort' returned exit status 0 while in /ssd/blink1/src/chrome/android/chrome_apk.gyp.

which makes sense because the customtabs is now moved as of change [0].


The GN build works fine, but run-webkit-tests is still busted.

Bots (% a few webrtc.fyi bots?) have switched to GN, including https://build.chromium.org/p/chromium.webkit/builders/Android%20Builder. I'm not surprised that android gyp has started rotting.

Me neither, but how is it not rotting for the Nexus 4 bot? Only me? 

If it is indeed rotting for everyone, including that bot, then I am blocked using run-webkit-tests on android. Can I get an update on the status of the fix for GN runs? Has anyone repro'd my failures?

The N4 bot doesn't build, it gets triggered by the builder I linked in #8.

Ah, sorry I missed that. Doesn't this suggest that it really should be possible for me to run-webkit-tests from a GN build? 

In an attempt to match the bot's command, I've changed my GN output directory to be out/Release -- still hitting the same failure. Can someone take a look this? I'm blocked. 

Deleted: out.txt 26.4 KB

out.txt 26.4 KB View Download

Can you capture the logcat for a run & upload that?

I'm suspicious of the use of FIFOs -- we had a lot of trouble with them in gtests on Android and moved away from them.

Here's the adb logcat. Its really short since this script dies so early/quickly.

Deleted: adb.log 4.8 KB

adb.log 4.8 KB View Download

Nothing particularly helpful in there, unfortunately.

I don't have access to my devices today, but I'll try to repro this tomorrow & get back to you.

Hey, just checking in

Sorry, busy week last week.

Looking at this now.

Layout tests work with gn. Still working on M+ support over in https://bugs.chromium.org/p/chromium/issues/detail?id=567947