On Sep 13, 2009, at 23:13:59 +0200 (CEST), Daniel Stenberg wrote:
How do these affect libcurl at all? Does the "control" part possibly
involve going over a proxy?
Unfortunately, I don't know how it's affecting libcurl. We're not
going over
a proxy
But how can it then affect libcurl at all?
We don't know. It has us absolutely stumped over here at the office.
I'm hoping someone had seen this before and was able to figure out
what was
going on.
Did you try using CURLOPT_DEBUGFUNCTION to get to see libcurl's idea
of what
it sends and compared that against what is actually seen on the wire?
Yes. libcurl thinks it's sending out the correct data.
Does this also happen if you just run "curl -d foobar [URL]" ?
Yes. The data still gets corrupted. I set up a cgi script on my
website so that anyone interested can try it out:
curl -d "Hello" http://www.malarkeysoftware.com/cgi-bin/echo.cgi
On a Mac OSX 10.5 admin or standard account, the resulting output is:
<html><body>
Hello
</body></html>
On the same machine but on an account with parental controls activated:
<html><body>
HelPr
</body></html>
Interestingly enough, I also tried constructing a raw HTTP POST
request and sending it via netcat to my server and it came through
just fine on the account with parental controls:
POST http://www.malarkeysoftware.com/cgi-bin/echo.cgi HTTP/1.1
Host: www.malarkeysoftware.com
Accept: */*
Connection: Keep-Alive
Content-Type: application/application/x-www-form-urlencoded
Content-Length: 5
Hello
> cat post_request.txt | nc www.malarkeysoftware.com 80
I tried to match as much as possible what libcurl was sending out.
Vince
