Project: chromium Issues People Development process History Sign in
New issue
Advanced search Search tips
Starred by 1 user
Status: Fixed
Closed: Apr 2013
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: ----

Sign in to add a comment
XP (dbg) bots are red, unknown cause
Reported by, Apr 13 2013 Back to list
All XP (dbg) tests are consistently failing starting at 193971-193975.

etc. Their release counterparts are green for the same revisions.

r193974 was reverted on the next build and does not seem to be the cause.
just checking, but were those system rebooted?

r193968 was also reverted and reapplied, busted'ness persisted throughout
First started in revision window: 193968 - 193975
The test targets fails to generate the list of tests contained within them.

> browser_tests.exe --lib=browser_tests --gtest_list_tests

... doesn't produce any output, so nofin happens.

It's not limited to just the browser_test target, other gunit test targets (all maybe) are affected too.
Labels: -Pri-0 Pri-1
Appears to be a sheriff issue, downgrading to P-1.
Some of the targets affected which might help find any code change (if it is a code change and not a system change) thats responsible.


... well, pretty much every test exe afaict. Without seeing a code change that's low enough to affect all of our test harnesses, makes me think its an environment thing somehow?
It's not all test.exes, its many or most, the set that fails to launch is consistent across runs.

base_unittest, ipc_tests, sql_unittests (amount some others) work for example
idk the cause, i'm leaving for the evening now
and making view unrestricted

This crept in on friday, bots were red dominic and i saw "test fail to launch mysteriously, thinking it was a bot'ism we pulled in xusydoc, and proceeded to do various other tree handling tasks with the xpbot'isms out of sight and out of mind for the time being. Early evening came around with no resolution so recorded what we know in this bug.

I can help poke at this on monday. Not clear wassup with some of the gtest based exes output on xp(dbg) bots? Are the binaries not producing output, or is the output being eaten somewhere so the scaffolding doesn't read it? Or are the dbg binaries crashing right at the start or what?

Logging into one of those systems and running...
> foo_tests.exe --lib=foo_tests --gtest_list_tests
... in a cmd window to see what happens may shed some light?
@xusydoc, did you try to manually run the binaries like that and if so what happened?

The way the listings are generated, the binaries go thru all the motions of running the tests, but then instead of running they just output the name of the test it would have run.

If no output is generated, what if instead of asking the binary to 'list_tests', ask it to run tests. I wonder if it'd run the tests w/o producing output?

The same dbg binaries produce output on other flavors of windows and run tests on other flavors of windows.

Anyway, this crept in on friday.
Labels: -Restrict-View-Google
Comment 11 by, Apr 13 2013
just a guess, but with the gpy[i] changes to rearrange gurllib, maybe we got a new system .dll dependency in dbg builds... and the .dll just isn't avail/installed on our xp bots... or some other change in the revwindow introduced such a dependency?
Status: Fixed
I reverted all googleurl -> url changes, and xp bots are happy again

I suspect that this is actually due to r193777 as this gave the following error:

url/url_canon_stdstring.h:65: error: undefined reference to 'vtable for url_canon::StdStringCanonOutput'

On the XP bots, the unit tests failed to load because of a missing import/export statements error.
Labels: -Build-Infrastructure Infra
Sign in to add a comment