Open GoogleCodeExporter opened 9 years ago
Original comment by andrewch...@chromium.org
on 30 Jan 2015 at 6:22
The rule is new specified "prefs" under "chromeOptions" will overwrite
existing
preferences, otherwise, it will not change it.
do you have test case that if we can reproduce it?
Original comment by andrewch...@chromium.org
on 30 Jan 2015 at 6:29
Thanks for the prompt feedback! I noted that when we specify "user-data-dir"
and the directory doesn't exist chromedriver creates a partial profile
directory between starting the session and invoking chrome. The issue is that
we use a special "binary" path that runs a script that copies one of our
template profiles (if not exists) before invoking chrome and it was complaining
that the directory already exists. In this case OK because we are now
overwriting any file. I will check if any preference is being changed when it
exists and get back to you, if not we can close the issue. In any case I think
chromedriver should not do anything if the "prefs" clause is not passed to
avoid unnecessary overhead.
Original comment by mbpp1...@gmail.com
on 31 Jan 2015 at 11:13
Let me how if everything work the way it should be.
Original comment by andrewch...@chromium.org
on 2 Feb 2015 at 6:22
So I did more tests and when the profile directory exists it seems no files are
changed by chromedriver or if changed it keeps the same content (I compared
some files before/after). You can close but note, when the profile doesn't
exist chromedriver is still creating a basic one even if we don't specify
"prefs". I think you should consider skipping this whole process if "prefs" is
not specified to save time and resources.
Original comment by mbpp1...@gmail.com
on 7 Feb 2015 at 2:46
If you don't specify "prefs", chromedriver will provide a default profile as
temp
this temp dir will be removed when session closed. you should not notice the
exist of temp dir.
Original comment by andrewch...@chromium.org
on 9 Feb 2015 at 6:51
Original comment by samu...@chromium.org
on 21 Feb 2015 at 12:18
But we use a custom data directory via "use-data-dir". In fact, before invoking
Chrome we copy our template to this data directory. We override the "binary"
parameter to point to a script that checks if the data directory exists before
invoking Chrome and at this point the directory is being pre-created by
Chromedriver even if we don't pass "prefs". Also, we see sometimes the error
"failed to write prefs file" (code 13), not sure if it is because we upgraded
Chromre recently (40.0.2214.115 m) and there is some profile mismatch or it is
related to this functionality.
Original comment by mbpp1...@gmail.com
on 23 Feb 2015 at 2:25
Getting in many tests the error:
unknown error: failed to write prefs file
I cannot determine any logical pattern, it happens in about 5% of the time and
no specific test or profile. One more reason not to touch the profile if no
"prefs" was specified. Versions:
Chromedriver: 2.14
Chrome 40
Original comment by mbpp1...@gmail.com
on 6 Mar 2015 at 5:48
Complementing, Windows 2008 R2 via Remote Desktop.
Original comment by mbpp1...@gmail.com
on 6 Mar 2015 at 5:49
since it happens randomly in about 5% of the time
1. can you do the same process in window 7(must use same chrome and
chromedriver)? - to see if it happen only in Windows 2008
2. Can you modify your scripts - copy prefs file before any write, and record
the status. compare the files when error
3. find a way to get OS detail messages about any write error - this is the
easiest
way to find out the problem.
Original comment by andrewch...@chromium.org
on 6 Mar 2015 at 8:02
In Windows 7 I was not able to see this error, I will try to get pre/post
versions of the preferences file in Windows 2008 when the error happens now. As
for OS errors I am not seeing any, nor in Windows Events. It seems to happen
when there is concurrency, when multiple tests are being executed under
different profiles (3+), my guess is that the error is related to some
wait/locking event in chromedriver. I will test more.
Original comment by mbpp1...@gmail.com
on 10 Mar 2015 at 2:26
some people use chromedriver in Selenium grid/node which test many devices
concurrently, and have not report the problem. In Windows 7, did you do the
same testing as in Window 2008?
Original comment by andrewch...@chromium.org
on 10 Mar 2015 at 6:17
I don't use Selenium, I implemented the protocol myself. I use chromedriver
running in a Remote Desktop session (not service). Then I start several tests
in parallel. I did the same test in Windows 7 and cannot replicate as I see in
Windows 2008. The test I did was to start 10 sessions at the same time. I still
cannot define a pattern or condition, it happens sometimes. Now there is only
one difference between the Windows 7 and Windows 2008 tests, the Chrome
version, one is 41 and the other 40. I will try to upgrade to do more tests.
Original comment by mbpp1...@gmail.com
on 11 Mar 2015 at 12:01
I upgraded to Chrome 41 and now I have another problem, not only "error writing
prefs file" but also some pages loading forever and I cannot even stop or open
any other tab. If a close the page Chrome continues to run on the background
and the only way to stop is to kill the process. Does it have to do with the
new DevTools enhancements in Chrome 41?
Original comment by mbpp1...@gmail.com
on 12 Mar 2015 at 8:15
So I downgraded to Chromediver 2.10, all tests executing fine with Chrome 41
without the "prefs" error and the freeze. I will keep one server with 2.14 and
41 to do more tests. The "freeze" error is specific to Chrome 41 and
Chromedriver 2.14. The "prefs" issue, would be nice if nothing is
done/written/overwritten if none is passed, if I can suggest as enhancement for
speed reasons and to avoid the "error writing prefs file" issues.
Original comment by mbpp1...@gmail.com
on 12 Mar 2015 at 10:40
I don't have Windows 2008. if you can focus in window 7 will be more helpful.
let's make sure testing for window 7 is clean.
Original comment by andrewch...@chromium.org
on 12 Mar 2015 at 11:45
I am still not able to simulate this in Windows 7, only Windows 2008 but I can
see that it happens when we execute 4+ parallel tests. The issue is that I
cannot yet identify some pattern, it can run fine in parallel for 30 minutes
and then suddenly it raises error code 13. What are the possible conditions for
this error? Since we use different profiles via "user-data-dir" it is clearly
not a file sharing issue. Can it be maybe some temporary file being generated
and there we have a sharing issue? In any case I will continue doing more
tests, I am now creating a custom script to simulate in Window 7 the same
environment workload we have in Windows 2008 (our production environment).
Original comment by mbpp1...@gmail.com
on 6 Apr 2015 at 7:06
Original issue reported on code.google.com by
mbpp1...@gmail.com
on 30 Jan 2015 at 11:12