Closed Bug 21083 Opened 25 years ago Closed 25 years ago

[dogfood]Setting multiple cookies at a time failing on M11

Categories

(Core :: Networking: Cookies, defect, P3)

Product:
Core
Component:
Networking: Cookies
Platform:
x86
Windows NT
Type:
defect

Tracking

()

Status:
VERIFIED FIXED

People

(Reporter: rxsherm, Assigned: morse)

References

(
URL
)

Details

If I login to my test site http://www.interealm.com/steel and use the ID "mozilla" with the password of "ummmm" the cgi stlogin.cgi sets up three cookies at the same time. They are: login = YES user= mozilla passwd= ummm Doing is in communicator 4.7 or IE works fine: All three cookies are set and the proper command screen thusly loads... but on M11, only first of the 3 cookies is set ("login=yes"), and so the login fails, throwing me into a "create new unit" type screen. It seems that although M11 takes my first cookie of the series allright, the others are not being accepted. If I change the order of the cookies being generated, M11 always takes the first one and the other two are ignored.
Sorry... forgot to add this is from build ID 1999120708
Correction: The password is "ummm" (3 m's) not "ummmm". Sorry about that.
Status: NEW → ASSIGNED
Summary: Setting multiple cookies at a time failing on M11 → [dogfood]Setting multiple cookies at a time failing on M11
Target Milestone: M12
Judson, last night you checked in a change to nsHTTPHeaderArray.cpp so that you are now separating concatentated cookies with carriage-return (\r) instead of line feed (\n). Why did you do that? Your comment says "replacing a char const w/ an int literal" but that doesn't tell me anything. Of course this broke cookie processing. Remember the code (that you added) to the SetCookieStringFromHTTP routine in nsCookie.cpp. There you test for '\n' in determining if there are concatenated cookies. :-( So we either need to back out the change that you just made (why did you make it?) or we need to change the code in nsCookie.cpp to test for '\r'. Which should I do (and why)?
OK, I spoke with Jud and I am going to modify his change. His intent was to replace '\n' with a literal for line feed but he accidentally used a literal for carriage return instead. So that's an easy fix and I will check it in just as soon as the tree goes green.
Status: ASSIGNED → RESOLVED
Closed: 25 years ago
Resolution: --- → FIXED
Fix is now checked in. Thanks, rxsherm, for catching and reporting this major regression so quickly.
verified: NT 2000012520
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.