791884 - talos now broken on windows mozharness...needs a better debug message

Status: RESOLVED FIXED

(Testing :: Talos, defect)

Description

[Jeff Hammel]

A week, perhaps, ago, I was able to run the jetperf on windows:

(mozharness) C:\Users\jhammel\mozharness\src\mozharness>python scripts\jetperf.p
y --config-file configs\jetperf\test_config.py --binary-path "c:\Program Files (
x86)\Mozilla Firefox\firefox.exe"

Now I get:

16:26:33    ERROR -  Traceback (most recent call last):
16:26:33     INFO -    File "C:\Users\jhammel\mozharness\src\mozharness\build\ve
nv\Scripts\talos-script.py", line 9, in <module>
16:26:33     INFO -      load_entry_point('talos==0.0', 'console_scripts', 'talo
s')()
16:26:33     INFO -    File "C:\Users\jhammel\mozharness\src\mozharness\build\ve
nv\lib\site-packages\talos\run_tests.py", line 295, in main
16:26:33     INFO -      run_tests(parser)
16:26:33     INFO -    File "C:\Users\jhammel\mozharness\src\mozharness\build\ve
nv\lib\site-packages\talos\run_tests.py", line 271, in run_tests
16:26:33     INFO -      talos_results.output(results_urls, **results_options)
16:26:33     INFO -    File "C:\Users\jhammel\mozharness\src\mozharness\build\ve
nv\lib\site-packages\talos\results.py", line 78, in output
16:26:33     INFO -      _output.output(results, url)
16:26:33     INFO -    File "C:\Users\jhammel\mozharness\src\mozharness\build\ve
nv\lib\site-packages\talos\output.py", line 67, in output
16:26:33     INFO -      f = file(results_path, 'w')
16:26:33     INFO -  IOError: [Errno 22] invalid mode ('w') or filename: ''
16:26:33    ERROR - Return code: 1
16:26:33    FATAL - Results file not found: C:\Users\jhammel\mozharness\src\mozh
arness\build\jetperf.txt
16:26:33    FATAL - Exiting -1

So I'm not sure what the problem is or what has changed.  I would like to add debugging output to talos to help figure this out.  Then we can figure out if this is a talos bug, a jetperf bug, or ...

Comment 1

[Jeff Hammel]

Attachment: more verbosity

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 662235 [details] [diff] [review]
more verbosity

Review of attachment 662235 [details] [diff] [review]:
-----------------------------------------------------------------

::: talos/output.py
@@ +70,5 @@
> +                    f.write(str(result))
> +                f.close()
> +            except Exception, e:
> +                print "results_url: %s" % results_url
> +                print "results_path: %s" % results_path

could we explain that there was an exception here?

Comment 3

[Jeff Hammel]

Attachment: a better (hopefully) error message

Comment 4

[Jeff Hammel]

pushed to try: https://tbpl.mozilla.org/?tree=Try&rev=8dca3c72221a

Comment 5

[Mozilla RelEng Bot]

Try run for 8dca3c72221a is complete.
Detailed breakdown of the results available here:
    https://tbpl.mozilla.org/?tree=Try&rev=8dca3c72221a
Results (out of 67 total builds):
    success: 67
Builds (or logs if builds failed) available at:
http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-8dca3c72221a

Comment 6

[Jeff Hammel]

pushed: http://hg.mozilla.org/build/talos/rev/38609147b9db

Comment 7

[Jeff Hammel]

Sorry, I didn't mean to close until I got more to the heart of this issue.  Muscle memory...

Comment 8

[Jeff Hammel]

So with this additional debugging I can see the following:

09:26:38     INFO -  Exception in writing file '' from results_url: file://C:\Us
ers\jhammel\mozharness\src\mozharness\build\jetperf.txt

not sure why we can't parse the bloody file out of there.

Comment 9

[Joel Maher ( :jmaher ) (UTC -8)]

a permissions issue maybe?  I am able to write results_url results to file:// locations on windows 7.

Comment 10

[Jeff Hammel]

No, its a parsing error, as we're not getting back for the path segment what we would think.  Will debug

Comment 11

[Jeff Hammel]

>>> import urlparse
>>> results_url = 'file://C:\Users\jhammel\mozharness\src\mozharness\build\jetpe
rf.txt'
>>> results_url_split = urlparse.urlsplit(results_url)
>>> results_url_split
SplitResult(scheme='file', netloc='C:\\Users\\jhammel\\mozharness\\src\\mozharne
ss\x08uild\\jetperf.txt', path='', query='', fragment='')

So apparently windows paths aren't parsed the same as unix paths o_O . Though now that I say that it does sound strikingly familiar.  In unix, the path goes in the (you guessed it) 'path' part of the segment just as God Intended.

:jmaher do you recall how you were able to avoid this error?

Sadly, I think the solution is to parse our own damn file:// urls.  This sounds like a job for mozfile.

Comment 12

[Jeff Hammel]

Now I don't even get that far:

C:\Users\jhammel\talos>Scripts\activate.bat
(talos) C:\Users\jhammel\talos>talos -n -d -a ts -e "c:\Program Files (x86)\Mozi
lla Firefox\firefox.exe" --develop --datazilla-url file://c:\Users\jhammel\foo.t
xt
setting debug
DEBUG: using testdate: 1348612496
DEBUG: actual date: 1348612496
RETURN:<a href = "http://hg.mozilla.org/releases/mozilla-release/rev/0b774a1067f
e">rev:0b774a1067fe</a>
qm-pxp01:
                Started Tue, 25 Sep 2012 15:34:56
Running test ts:
                Started Tue, 25 Sep 2012 15:34:56
DEBUG: operating with platform_type : w7_
DEBUG: created profile
        Screen width/height:1600/900
        colorDepth:24
        Browser inner width/height: 1010/675


browser_name:Firefox
browser_version:15.0.1
buildID:20120905151427

DEBUG: initialized firefox.exe
DEBUG: command line: '"c:\Program Files (x86)\Mozilla Firefox\firefox.exe"  -pro
file c:\\\users\\\jhammel\\\appdata\\\local\\\temp\\\tmp0r5hy1\\\profile http://
localhost:15707/startup_test/startup_test.html?begin='
DEBUG: unknown error during cleanup
Traceback (most recent call last):
  File "C:\Users\jhammel\talos\Scripts\talos-script.py", line 8, in <module>
    load_entry_point('talos==0.0', 'console_scripts', 'talos')()
  File "C:\Users\jhammel\talos\talos\run_tests.py", line 295, in main
    run_tests(parser)
  File "C:\Users\jhammel\talos\talos\run_tests.py", line 250, in run_tests
    talos_results.add(mytest.runTest(browser_config, test))
  File "C:\Users\jhammel\talos\talos\ttest.py", line 417, in runTest
    self.cleanupProfile(temp_dir)
  File "C:\Users\jhammel\talos\talos\ttest.py", line 147, in cleanupProfile
    self._hostproc.removeDirectory(dir)
  File "C:\Users\jhammel\talos\talos\ffprocess.py", line 142, in removeDirectory

    shutil.rmtree(dir)
  File "C:\Python27\Lib\shutil.py", line 245, in rmtree
    rmtree(fullname, ignore_errors, onerror)
  File "C:\Python27\Lib\shutil.py", line 245, in rmtree
    rmtree(fullname, ignore_errors, onerror)
  File "C:\Python27\Lib\shutil.py", line 250, in rmtree
    onerror(os.remove, fullname, sys.exc_info())
  File "C:\Python27\Lib\shutil.py", line 248, in rmtree
    os.remove(fullname)
WindowsError: [Error 32] The process cannot access the file because it is being
used by another process: 'c:\\users\\jhammel\\appdata\\local\\temp\\tmp0r5hy1\\p
rofile\\Cache\\_CACHE_001_'

(talos) C:\Users\jhammel\talos>

Comment 13

[Jeff Hammel]

In the above case, the browser never closed.  Not sure why this is.  Worth making sure the startup test works at all on windows

Comment 14

[Jeff Hammel]

Confirmed that I can in fact run tsvg now.  If it ain't one thing, its another thing

Comment 15

[Jeff Hammel]

(In reply to Jeff Hammel [:jhammel] from comment #13)
> In the above case, the browser never closed.  Not sure why this is.  Worth
> making sure the startup test works at all on windows

This is actually bug 794678 and unrelated to this bug.

Comment 16

[Jeff Hammel]

This now works