HTTP sampler corrupts filename with Chinese characters when upload file with http sampler

[asfimport]

Nick.Yin (Bug 61669):
precondition：
use multipart/form-data for Post and Browser-compatible headers is chosen.

step:
1.add a HTTP sample to threadgroup
2.choose a file with Chinese character in file upload area in sampler
3. excute the test

excepted result：
the file whose file name is Chinese character is displayed in view result tree

actual:
the file name displays like ?????

by the way: it was working fine with Jmeter 3.1

Created attachment filename1.PNG: filename with chinese charactor

Severity: major
OS: All

Duplicates:

• HTTP request unicode file name #4699
• Jmeter can't upload Chinese files as attachment, has encoding problem #5144
• File upload Content-Disposition encoding error #5529

[asfimport]

Nick.Yin (migrated from Bugzilla):
Created attachment filename2.PNG: filename with English

[asfimport]

@pmouawad (migrated from Bugzilla):
See:

• https://stackoverflow.com/questions/20591599/why-arent-post-names-with-unicode-sent-correctly-when-using-multipart-form-data/20592910#20592910

RFC 2388 4.4 applies here and the filenames can be encoded using RFC 2231.

See discussion around #4460

[asfimport]

Nick.Yin (migrated from Bugzilla):
Created attachment http-request with jmeter3.2.png: http-request of jmeter3.2

[asfimport]

Nick.Yin (migrated from Bugzilla):
Created attachment result of jmeter3.2.png: result of jmeter3.2

[asfimport]

Nick.Yin (migrated from Bugzilla):
Created attachment http-request with jmeter3.3.png: http-request of jmeter3.3

[asfimport]

Nick.Yin (migrated from Bugzilla):
Created attachment result of jemeter3.3.png: result of jmeter3.3

[asfimport]

Nick.Yin (migrated from Bugzilla):
I tried with Jmeter 3.2 compared to Jmeter 3.3, and found it's working fine with Jmeter 3.2. See the attachment.
and I reviewed the release note of Jmeter3.3 , I'm still thinking it's a bug.(In reply to Philippe Mouawad from comment 2)

See:

https://stackoverflow.com/questions/20591599/why-arent-post-names-with-
unicode-sent-correctly-when-using-multipart-form-data/20592910#20592910

RFC 2388 4.4 applies here and the filenames can be encoded using RFC 2231.

See discussion around #4460

[asfimport]

@pmouawad (migrated from Bugzilla):
Hello,
It's not a regression although you might think so.

JMeter 3.2 was wrong in working the way it was.
The change is due to #4460.

See :

• http://httpcomponents.10934.n7.nabble.com/Charset-restriction-on-parameter-names-td31380.html
• https://issues.apache.org/jira/browse/HTTPCLIENT-293
• Discussion with title:
  svn commit: r1806215 - in /jmeter/trunk: src/protocol/http/org/apache/jmeter/protocol/http/sampler/HTTPHC4Impl.java xdocs/changes.xml

Also have a look at TEST_HTTP.jmx you'll see that tests consider this.

If you think I am wrong in my analysis, can you point me to the RFC that contradicts JMeter behaviour ?

Thank you

[vlsi]

The syntax is to send the filename twice: https://greenbytes.de/tech/webdav/rfc5987.html#examples

We probably need implementing it

[user412857]

The syntax is to send the filename twice: https://greenbytes.de/tech/webdav/rfc5987.html#examples

We probably need implementing it

Can you be more precise? How to send the filename twice? How to do that using BeanShell PreProcessor instead of HTTP Request Files Upload?
Maybe a sample would help

[vlsi]

Let me explain:

1. As seen in HTTP sampler corrupts filename with Chinese characters when upload file with http sampler #4546 (comment), JMeter uses Content-Disposition: form-data header to pass files.
2. However, historically, most headers supported ASCII characters only. Currently, RFC 7578 suggest a way out to pass unicode filenames by percent-encoding the name.

So I would suggest to try adjusting JMeter code that creates Content-Disposition to match RFC 7578:

jmeter/src/protocol/http/src/main/java/org/apache/jmeter/protocol/http/sampler/PostWriter.java

Line 391 in 88423ed

write(out, "Content-Disposition: form-data; name=\""); // $NON-NLS-1$

.

Do you think you could create such a pull request?

[vlsi]

For reference, HttpClient 4.x does not support Unicode filenames, and https://issues.apache.org/jira/browse/HTTPCLIENT-293 was integrated into 5.x only.

[vlsi]

I've prepared a fix in #5987.
It would be great if someone could test it with their systems.