| Ticket UUID: | 2a94652ee10cae20c0b54cf9fd6516eb01daadd5 | ||
| Title: | http package reply "100 continue" code broken | ||
| Type: | Bug | Created on: | 2017-08-30 14:23:17 |
| Submitter: | oehhar | Assigned to: | oehhar |
| Subsystem: | 29. http Package | Severity: | Critical |
| Priority: | 5 Medium | Last modified: | 2017-08-31 10:03:56 |
| Status: | Closed | Closed by: | oehhar |
| Resolution: | Fixed | Closed on: | 2017-08-31 10:03:56 |
| Version: | 8.6.7 | ||
| Description: | ||||
|
http 2.8.11 package does not correctly ignore "100 Continue" replies. The relevant code is broken. What happens? https://tools.ietf.org/html/rfc7231#section-6.2.1 The first line of a web server reply contains the status. If this is: HTTP/1.1 100 Continethe real header will follow as a following line. The current code correctly ignores the continue status, but does not expect a following status line. The insertion of a set to state "connecting" fixes that (around line 1034):
proc http::Event {sock token} {
...
# We ignore HTTP/1.1 100 Continue returns. RFC2616 sec 8.2.3
if {$state(http) == "" || ([regexp {^\S+\s(\d+)} $state(http) {} x] && $x == 100)} {
set state(state) "connecting"
return
}
At least, inserting this line cured the communication for me with TCLWS and a .net server. | ||||
| User Comments: | ||||
oehhar added on 2017-08-31 06:35:22:
Branch [bug-2a94652ee1] with commit [b80bf5033f] inserts the proposed line and increses the version number of the http number. I would appreciate a review before merge or just a merge. The fix is IMHO ok and works well for me. The review is mor for formal issues like version number increment and targeting branches. jan.nijtmans added on 2017-08-31 08:56:30:
Well, I cannot find anything wrong with it. So, just go ahead (with merging to core-8-6-branch -> trunk -> novem)! Thanks! oehhar added on 2017-08-31 10:03:56:
Ok, committed to all branches. Thank you ! | ||||