Closed GoogleCodeExporter closed 8 years ago
Thanks for the bug report and screenshot. I think I have fixed the bug in r985
- can you download 1.3.1-RC1 to
see if it works properly for you?
http://namebench.googlecode.com/files/namebench-1.3.1-RC1-Windows.exe
Thanks!
Original comment by helixblue
on 5 Jun 2010 at 5:03
I got another error but this one is probably a typo in the script.
Original comment by VinceVri...@gmail.com
on 5 Jun 2010 at 5:45
Attachments:
Sorry about that. Does 1.3.1-RC2 work any better?
http://code.google.com/p/namebench/downloads/list
Original comment by helixblue
on 5 Jun 2010 at 6:19
I just tried RC3 as that was the latest one when i looked. It did work without
raising an error now. I got odd results tough. Everything timed out on every
DNS. I
attached the results html page. The uploaded public result page
(http://namebench.appspot.com/id/553005) also spits out the following python
error:
Traceback (most recent call last):
File
"/base/python_runtime/python_lib/versions/1/google/appengine/ext/webapp/__init__
.py",
line 511, in __call__
handler.get(*groups)
File "/base/data/home/apps/namebench/3.342383898571905264/lookup.py", line 71, in get
'min_duration_url': self._CreateMinimumDurationUrl(nsdata, key="min-%s" % id),
File "/base/data/home/apps/namebench/3.342383898571905264/lookup.py", line 123, in
_CreateMinimumDurationUrl
url = charts.MinimumDurationBarGraph(min_data)
File "/base/data/home/apps/namebench/3.342383898571905264/libnamebench/charts.py",
line 131, in MinimumDurationBarGraph
labels = range(0, scale, tick) + [scale]
TypeError: range() integer step argument expected, got NoneType.
Original comment by VinceVri...@gmail.com
on 5 Jun 2010 at 10:50
Attachments:
Given your earlier error, I did expect some strange results, but nothing like
this. I've updated the
namebench.appspot.com website with the same code so that it no longer crashes
as well. It may seem
unbelievable, but the results show that Python thinks the clock on your
computer is moving backwards
sometimes rather than forwards.
Does your computer have multiple cores or processors? If so, I bet we are
running into this quirk in Windows:
http://www.virtualdub.org/blog/pivot/entry.php?id=106
Can you run this command from a command-line for me - I am curious what the
output looks like:
ping 192.168.1.1
I may try to put a workaround in place that we detect this kind of issue, and
switch to a less accurate time
source. Thinking it through now.
Original comment by helixblue
on 6 Jun 2010 at 6:55
I've implemented a workaround in 1.3.1-RC4 that I would like you to test if
possible:
http://code.google.com/p/namebench/downloads/list
I've added some detection to see if the high-precision clock appears to be
going backwards, and if so, switch to a
lower precision version. I'm not fully certain this technique works well, so I
am eager to see what it does for you.
Original comment by helixblue
on 6 Jun 2010 at 7:37
Sometimes it returns a negative ping, i got: -12ms, 14ms, 15ms and -15ms. I
looked it
up a while ago and it's because of my AMD processor. I have an AMD Athlon X2
3800.
I just tried RC4, got another error, see the attachment. I will be leaving this
computer in 3 hours and i will not come back before 2 weeks, so i won't be able
to
test for a while.
Original comment by VinceVri...@gmail.com
on 6 Jun 2010 at 9:20
Attachments:
Thanks again for the feedback! I've posted RC5, which tries to improve the
detection of whether to use
time.clock() or time.time() by inserting micro-sleeps in between calls. Once it
has selected an appropriate
clock to use, if it ever encounters a negative time, it will crash loudly with:
BrokenSystemClock('The time on your machine appears to be going backwards. '
'We cannot accurately benchmark due to this error. '
'(timer=%s, duration=%s)' % (self.timer, duration))
If you get a chance, I'm curious to see if it makes any difference. If you get
the BrokenSystemClock message, I
am curious to see which timer it says it is using.
From further research, I found http://wiki.ljackson.us/Negative_Ping_Time which
mentions that AMD posted a
fix for this issue for Windows XP and Windows 2003 users. To download it, go to:
http://www.amd.com/us-en/Processors/TechnicalResources/0,,30[_]182[_]871[_]13118
,00.html
Search the page for: AMD Athlon™ 64 X2 Dual Core Processor Driver for Windows
XP and Windows Server 2003
Version (x86 and x64 exe) 1.3.2.16
Original comment by helixblue
on 6 Jun 2010 at 10:30
I just tried RC5 WITHOUT the processor fix. This is the reulst page. I have to
go
now, I'll try the fix in 2 weeks.
Original comment by VinceVri...@gmail.com
on 6 Jun 2010 at 11:47
This should be fixed as of 1.3.1. It will now complain if time goes backwards
consistently.
Original comment by helixblue
on 15 Jul 2010 at 1:15
Original issue reported on code.google.com by
VinceVri...@gmail.com
on 5 Jun 2010 at 4:22Attachments: