You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@jmeter.apache.org by bu...@apache.org on 2016/12/05 21:34:54 UTC
[Bug 60444] New: Intermittend failure of
TestHTTPMirrorThread#testSleep()
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
Bug ID: 60444
Summary: Intermittend failure of
TestHTTPMirrorThread#testSleep()
Product: JMeter
Version: 3.0
Hardware: PC
Status: NEW
Severity: minor
Priority: P2
Component: HTTP
Assignee: issues@jmeter.apache.org
Reporter: ts-nospam12@online.de
Target Milestone: ---
Created attachment 34500
--> https://bz.apache.org/bugzilla/attachment.cgi?id=34500&action=edit
complete log of ant build
This might be OS specific: Running the tests, I'm occasionally stalled by the
following test failure:
[java] There was 1 failure:
[java] 1)
testSleep(org.apache.jmeter.protocol.http.control.TestHTTPMirrorThread)
[java] junit.framework.AssertionFailedError: Expected > 1000 992
[java] at junit.framework.Assert.fail(Assert.java:57)
[java] at junit.framework.Assert.assertTrue(Assert.java:22)
[java] at junit.framework.TestCase.assertTrue(TestCase.java:192)
[java] at
org.apache.jmeter.protocol.http.control.TestHTTPMirrorThread.testSleep(TestHTTPMirrorThread.java:413)
I'm not too certain about the primary cause, which I think, is one of the
following:
a.) Thread.sleep() relies on the granularity of the internal clock, which might
shave off some millis from the 1000 ms, that have been asked for.
b.) The JVM generally isn't guaranteed to wait the full time given anyway, as
the sleep might get interrupted.
Now I'm a bit at a loss, what is the intention here:
If accuracy in maintaining the contract is required, I may fix this, by
tracking the actual time spent waiting arround the invocation of
TimeUnit.MILLISECONDS.sleep(Integer.parseInt(sleepHeaderValue))
@ HttpMirrorThread:223.
else we might just relax the assertion wihin the test.
Do we ever need this functionality?
Wadayathink?
P.S.:
The issue is annoying, since it breaks the test execution in one out of three
executions on my machine, and it did so since I startet to dig through the code
with Rel 3.0
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
--- Comment #6 from Thomas Schapitz <ts...@online.de> ---
Created attachment 34542
--> https://bz.apache.org/bugzilla/attachment.cgi?id=34542&action=edit
Patch as promissed.
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
--- Comment #3 from Thomas Schapitz <ts...@online.de> ---
If checked up on my previous statement, and have to correct:
connect() actually does initiate the roundtrip to the server, at least for the
Sun implementation of the HttpUrlConnection.
Thus, we should indeed start taking time before the connect(). I'm going to
supply a patch for this, which should mitigate the woes.
However I would like to point out that this still isn't a healthy aproach:
sleep() throws an InterruptedException, which is apart from a log statement
silently swallowed, so the test does hinge on the assumption, that there are no
spourious interruptions.
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
--- Comment #5 from Thomas Schapitz <ts...@online.de> ---
@Felix: I'm, already shaving the yak here - my intention was just to get the
tests going again.
Introducing additional erromessages, albeit sensible, invites compatibility
trouble down the line.
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
Thomas Schapitz <ts...@online.de> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #34500|0 |1
is obsolete| |
--- Comment #2 from Thomas Schapitz <ts...@online.de> ---
Created attachment 34525
--> https://bz.apache.org/bugzilla/attachment.cgi?id=34525&action=edit
ant log
Sorry about the log, there it is now.
Anyway, it's only puting the stacktrace into some context. This happens on a
Windows 10 Notebook, both when running with JDK 1.7.0 or 1.8.0.
From my point of view, the current position of now = System.nanoTime() is quite
right. Placing it earlier is equivalent of adding some time, since the server
side method will not get executed, until either getInputStream() or
getOutputStream() are called, *and* the streams are read from or written to -
thats at least what I took from various internet posts about the
HttpUrlConnection. You might want to convince yourself by placing breakpoints
into the test and the serverside method.
My questions are on a different tack:
- What are we trying to achieve by the test?
which also relates to the question:
- what is the use case of a server side wait in a JMeter HTTP-Mirror?
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
--- Comment #4 from Felix Schumacher <fe...@internetallee.de> ---
Thanks for your further analysis.
Perhaps let the server return a result code/string, that could be parsed in the
test method and checked for an (Interrupted)Exception?
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittent failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
Philippe Mouawad <p....@ubik-ingenierie.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|Intermittend failure of |Intermittent failure of
|TestHTTPMirrorThread#testSl |TestHTTPMirrorThread#testSl
|eep() |eep()
CC| |p.mouawad@ubik-ingenierie.c
| |om
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittend failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
Felix Schumacher <fe...@internetallee.de> changed:
What |Removed |Added
----------------------------------------------------------------------------
OS| |All
Status|NEW |NEEDINFO
--- Comment #1 from Felix Schumacher <fe...@internetallee.de> ---
On what computer do you see these failures?
I think it might help to put the now=System.nanoTime() before the
conn.connect().
And by the way the attached log is only one byte long.
--
You are receiving this mail because:
You are the assignee for the bug.
[Bug 60444] Intermittent failure of TestHTTPMirrorThread#testSleep()
Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=60444
Philippe Mouawad <p....@ubik-ingenierie.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |FIXED
Status|NEEDINFO |RESOLVED
--- Comment #7 from Philippe Mouawad <p....@ubik-ingenierie.com> ---
Thanks for contribution.
Date: Wed Dec 21 15:25:34 2016
New Revision: 1775470
URL: http://svn.apache.org/viewvc?rev=1775470&view=rev
Log:
Bug 60444 - Intermittent failure of TestHTTPMirrorThread#testSleep()
Bugzilla Id: 60444
Modified:
jmeter/trunk/test/src/org/apache/jmeter/protocol/http/control/TestHTTPMirrorThread.java
jmeter/trunk/xdocs/changes.xml
--
You are receiving this mail because:
You are the assignee for the bug.