Help: More stuck WebKit builders

[Stephen White]

WebKit Win7 and WebKit Win7 (dbg) (2) seem to be stuck trying to kill the webkit_tests step; could someone do what's necessary please?

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7%20%28dbg%29%282%29

Stephen

[Peter Schmidt]

On Wed, Oct 31, 2012 at 9:56 AM, Stephen White <senor...@google.com> wrote:

WebKit Win7 and WebKit Win7 (dbg) (2) seem to be stuck trying to kill the webkit_tests step; could someone do what's necessary please?

Done.. 

[Stephen White]

On Wed, Oct 31, 2012 at 1:02 PM, Peter Schmidt <psch...@google.com> wrote:

On Wed, Oct 31, 2012 at 9:56 AM, Stephen White <senor...@google.com> wrote:

WebKit Win7 and WebKit Win7 (dbg) (2) seem to be stuck trying to kill the webkit_tests step; could someone do what's necessary please?

Done.. 

Thanks.  Unfortunately, even after a reboot, the same bots (plus the WebKit XP bot) seem to be getting into the same state:

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20XP/builds/413

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7/builds/519

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7%20%28dbg%29%282%29/builds/375

It looks like the webkit tests finish correctly, but the system blocks trying to kill the process.

Stephen

[Chase Phillips]

When this happened last time, it was due to code changes in WebKit/Chromium.  In short: we can force a reboot on the system, but because something in the test harness or test code changed to cause the problem in the first place, it will happen again in the next cycle.

I'll take a look to see if anything's obviously wrong.  If someone knows of recent changes to the webkit tests step code, speak up.  Such changes could be related.

[Stephen White]

I'm trying to track it down to a specific revision, but it doesn't look like it corresponds to a WebKit or Chrome rev.

last known good win webkit build:  chrome r165045 wk r132885

last known good win webkit build:  chrome r165056 wk r132955

first known bad win webkit build:  chrome r165062 wk r132891

first known bad win webkit build:  chrome r165065 wk r132965

first known bad win webkit build:  chrome r164894 wk r132893

(Note that this is an agglomeration of DEPS and canary results, so it's a bit confusing).

So it looks like one bot failed  at Chrome 164895, but another bot passed at 165045.

One bot failed at WebKit r132891, but another bot passed at 132955.

So I don't think it's a WebKit or Chrome change.

(And once a bot starts hanging, it continues hanging after a reboot, so it's not a flakiness problem AFAICT).

Stephen

[Dirk Pranke]

I'm helping w/ this now, for whatever it's worth :).

[Dirk Pranke]

I believe I've identified a bug in NRWT that was keeping us from
killing DRTs when they timed out, and so we weren't cleaning up
properly. I've landed a fix for this in r133076, so let me know if you
see any problems after that.

-- Dirk

[Bev Cristobal]

Hello, 

Could the win_layout  try bots be experiencing a similar problem:

http://build.chromium.org/p/tryserver.chromium/builders/win_layout/builds/983

http://build.chromium.org/p/tryserver.chromium/builders/win_layout_rel/builds/894

Thank you,

Bev 

[Dirk Pranke]

Yes. Those builds show the bots at webkit r133029, which is before my
fix went in.

That's really old; anyone know why are the try servers that out of date?

-- Dirk

[Wez]

What base revision of WebKit do the bots use for WebKit patch tries?  Do they try to match the patch base revision, or the WebKit dep revision of Chrome's LKGR, or something else?

[Dirk Pranke]

I believe by default it'll pick up whatever version of WebKit is
referenced in the LKGR DEPS file.

-- Dirk