Gerry, Please submit a patch for HTTPSender/CommonsHTTPSender ASAP!. Don't worry about a test case right now...i'd like to review the code patch first.
thanks, dims On 6/13/05, Gerry Gao <[EMAIL PROTECTED]> wrote: > Hi dims, > > Well done! And hope this is not too late. > > I still have some concern about http://issues.apache.org/jira/browse/AXIS-895. > > As I reviewed the latest CVS code and do a basic test on it, I think there is > still some issue will the cookie stuff in HTTPSender and CommonsHTTPSender. > > Now the HTTPSender and CommonsHTTPSender have different cookie behavior for > "maintainSession" client, and both of the behaviors are not completely > correct. > > According to my test, now HTTPSender is a closer to the perfect. > > Here is a sample for HTTPSender now (paste from tcpmon): > > request 1: > <no cookie> > respones 1: > Set-Cookie: JSESSIONID=14qbbc5mptua2;Path=/axis > Set-Cookie: a=a > Set-Cookie: b=b > > request 2: > Cookie: JSESSIONID=14qbbc5mptua2 > Cookie: a=a > Cookie: b=b > respnonse 2: > Set-Cookie: a=a1 > Set-Cookie: c=c > > request 3: > Cookie: JSESSIONID=14qbbc5mptua2 > Cookie: a=a > Cookie: b=b > Cookie: a=a1 > Cookie: c=c > > IMO, the reqesut 3 should only have one cookie with name "a" and it should be > "a=a1" instead of tow cookies. > > For the same sitation for CommonsHTTPSender, it can't even maintain the > session correctly! > > request 1: > <no cookie> > respones 1: > Set-Cookie: JSESSIONID=bpg1ne69gl099;Path=/axis > Set-Cookie: a=a > Set-Cookie: b=b > > request 2: > Cookie: JSESSIONID=bpg1ne69gl099 > Cookie: a=a > Cookie: b=b > respnonse 2: > Set-Cookie: a=a1 > Set-Cookie: c=c > > > request 3: > Cookie: a=a1 > Cookie: c=c > > With CommonsHTTPSender, it only send to sever the cookies which are added in > the very last response. > > This result is consistent with the code I reviewed. > > IMO, it should use HashMap or Hashtable instead of String[] to handle > multicookies. > Hope this helps. > > BTW, I do want submit a test case to reoccur this, but I don't know if it > must be a standalone runnable JUnit test case. Actually, don't know how to > pack a test comfortable for AXIS. > > Gerry > > ----- Original Message ----- > From: "Davanum Srinivas" <[EMAIL PROTECTED]> > To: <[email protected]>; <[email protected]> > Sent: Monday, June 13, 2005 11:03 AM > Subject: Re: Blockers for Axis 1.2.1 Release(?) > > > Team, Folks, > > Zero blockers now...Please try latest CVS / Nightly ASAP! > > thanks, > dims > -- Davanum Srinivas -http://blogs.cocoondev.org/dims/
