On 13/09/06, Andreas Krennmair <[EMAIL PROTECTED]> wrote:
sebb schrieb:
> On 12/09/06, Andreas Krennmair <[EMAIL PROTECTED]> wrote:
>> Hello everybody,
>>
>> Currently, I use the "HTTP Request" sampler to make HTTP requests, with
>> the "Retrieve all Embedded Resource from HTML Files" option enabled.
>>
>> What I noticed when I looked at the requests in the results tree that
>> the Host headers seem to be wrong. When I request
>> http://mo-int.somedomain.com/web/download.do?method=showTrack&orderId=1234567
>>
>> which contains an image with the URL
>> http://img-int.somedomain.com/img/app?path=98765432.jpg&width=157&height=157
>>
>> then this image is requested, but in this HTTP request, the HTTP "Host"
>> header is set to "mo-int.somedomain.com", i.e. the host name of the page
>> into which it is actually embedded.
>>
>> Can anybody else reproduce this? If so, it should be fixed, IMHO, as it
>> is a violation of RFC 2616, section 14.23. BTW, I use JMeter 2.2, but
>> had the very same problem with previous versions.
>
> Sounds like a bug; I suggest you create a Bugzilla issue.
>
> Does it happen with both HTTP Samplers?
>
> Is there a public web-site you know of that can be used to demonstrate
> the problem?
I just hacked together a small example:
http://synflood.at/tmp/demo.html
(notice synflood.at vs. www.synflood.at in the URLs!)
And here's a test plan that triggers this issue:
http://synflood.at/tmp/bugdemo.jmx
The problems seems to be that browser-derived headers (I recorded this
using the HTTP Recording Proxy) actually overrule not only the headers
for the request of the HTML file, but also for the embedded files that
are downloaded.
It now becomes clear.
If you use the proxy to record a test plan, it is not normally
necessary to retrieve embedded resources, as that will have been done
by the browser.
Having said that, JMeter should probably not save the Host headers, as
these are generated by the lower level protocol.
Just delete the Host headers from the appropriate config records, and
it should work OK.
Can you verify this, now that I have produced a demo page plus a test
plan with which I am able to reproduce this behaviour?
Regards,
Andreas Krennmair
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
