jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: svn commit: r1511681 - in /jmeter/trunk: src/protocol/http/org/apache/jmeter/protocol/http/sampler/HTTPAbstractImpl.java xdocs/usermanual/component_reference.xml
Date Fri, 09 Aug 2013 17:19:17 GMT
On 9 August 2013 10:43, Milamber <milamber@apache.org> wrote:
>
> Le 08/08/2013 20:43, sebb a ecrit :
>>
>> [snip]
>>
>> -    private static final String DEVICE_PREFIX = "/dev/";
>> +    private static final String DEVICE_PREFIX = "/";
>>>>>
>>>>>
>>>>>
>>>>> This changes introduce an regression, /dev/eth0 don't works now.
>>>>>
>>>> Yes, that was deliberate. I changed the docs accordingly.
>>>>
>>>> You need to use /eth0.
>>>
>>>
>>> /eth0 don't works, but /ipv4/eth0 works.
>>
>> That should work - it works for me on Win/XP.
>>
>> Add some debug and see why it's not working.
>>
>>> Seems very complicated to find the good syntax (without read the docs or
>>> with "IP source address" label only)
>>
>> Where else apart should it be described?
>>
>>> Why not use a regexp pattern to check IPv4 and IPv6 address? without a
>>> ipvX
>>> prefix ?
>>
>> Not sure I understand.
>>
>>> And why not considering if the ipSource (as is) isn't a IP address (4/6),
>>> and not is in the interface's list on host, then it's a hostname, else
>>> return an error.
>>> Therefore it's not necessary to have special prefix to fill the field.
>>
>> Two issues:
>> - if the interface name is checked first, it will override the
>> identical hostname, which could cause existing tests to fail (not all
>> that likely, but possible)
>> - if the name is not an interface, the check is unnecessary
>>
>> That's why I chose a prefix that cannot be present in a host name.
>
>
> Ok, in this case, perhaps, a (better?) solution is to provide a combo box to
> allow the user to choice the source address field.

I was trying to avoid increasing the complexity of the GUI.

It's a fairly unusual use-case, so I took the view that the user would
not mind putting up with a slightly unusual syntax.

> With the combo box, the user could use the common / usual name for their
> device (only eth0, wlan0, etc.), IP or hostname (short or fdqn)
>
> Here a screenshot of HTTP Request with some improvements for reduce the
> width screen size (font changes for the Optional panel and HTTP options) +
> the Src IP combo box.
> http://www.milamberspace.net/img/http-request-src-addr.png

The IPv4/6 addr entries are ambiguous.
Do they apply only to devices?
I would hope so, otherwise existing test plans will break.
In which case, the entries need to be renamed.

The field currently supports Hostname or IP address; it is important
that compatibilty is maintained.

So the first entry should be for the Hostname/IP addr.
The entries need to be something like:

Hostname/IP
Device
Device IPv4
Device IPv6

> I can commit the UI changes for reduce the minimal width of the HTTP
> Request's pane,

Although the GUI is roughly the same width, it is more complicated.

> and if your are agree the changes to add the combo box?

I'm not convinced the change makes it easier for the user.
There would be two fields to configure instead of one.

The fields should be enclosed in a box so that it is clear they are related.
For example, as is done for Web Server & Timeouts.

Maybe the "Embedded URLs must match" field should also have a border.

> (please note, the rename of Source IP address field)

That's OK.

>
>
>
>
>>
>>>
>>>
>>>
>>>> I thought it was awkard to use /dev/ipv6/eth0, so I collapsed the prefix
>>>> to "/".
>>>>
>>>>>> +    private static final String IPV4 = "ipv4/";
>>>>>>
> [snip]

Mime
View raw message