Received: (qmail 17939 invoked by uid 2012); 21 Oct 1997 16:25:44 -0000 Message-Id: <19971021162544.17938.qmail@hyperreal.org> Date: 21 Oct 1997 16:25:44 -0000 From: Lee Pollington Reply-To: webmaster@cursci.co.uk To: apbugs@hyperreal.org Subject: post method randomly works X-Send-Pr-Version: 3.2 >Number: 1292 >Category: mod_cgi >Synopsis: post method randomly works >Confidential: no >Severity: critical >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Tue Oct 21 09:30:00 PDT 1997 >Last-Modified: Wed Oct 22 10:30:00 PDT 1997 >Originator: webmaster@cursci.co.uk >Organization: >Release: 1.2.4 >Environment: SunOS Electrop 5.5.1 Generic_103640-04 sun4m sparc SUNW,SPARCstation-20 >Description: Sporadic delivery of cgi data sent through the post method, sometimes get the data on reload of browser. >How-To-Repeat: No url available as I have had to go back to previous version of apache. >Fix: here's anextract I found on news:...servers/unix/misc: Subject: Re: Failed CGI From: Haseeb Asghar Date: 1997/10/16 Message-Id: <3445FD3E.304C@gssec.bt.co.uk> Newsgroups: comp.infosystems.www.servers.unix [More Headers] Steve Rickman wrote: > > This problem seems to be caused by a miscommunication between Apache > 1.2.4 and certain browsers, builds of Navigator 3.03 among them. It > occurs intermittently when CGI scripts are requested with method POST. > The problem may be related to the FIN_WAIT_2 glitch discussed on the > apache.org site at http://www.apache.org/docs/misc/fin_wait_2.html. > The only fix I know of is to use method GET instead of method POST. > > The problem has been reported by others on several groups, but it > remains little known, presumably because users just shrug it off as a > slow server or buggy CGI script. > > I have seen no firm list of what browsers are affected. I do know it > to be a problem in Navigator 3.03 (Win95, I). I have verified that it > does not occur with Navigator 3.01 (16-bit, I) or with MSIE 3.02 for > Win95. I have been informed that it is not a problem in Navigator 4.x. > It would probably be helpful if others could report results with > specific browsers (including both version number and platform, of > course). > Im having a similar problem... when i run cgi scripts via apache 1.2.4 theyre crashing my navigator 3.02 browser... but it isnt consistent..sometimes the scripts run fine.. other times they crash the browser... someone suggested that it may be because the scripts are returning http 1.1 headers to the browser.. and the browser is only compatible with http 1.0 headers?? any thoughts??? Note: netscape gold..and internet explorer dont crash.. they run the scripts fine...all the time Hasee >Audit-Trail: State-Changed-From-To: open-closed State-Changed-By: dgaudet State-Changed-When: Tue Oct 21 17:38:31 PDT 1997 State-Changed-Why: This is probably the same problem as that described in PR#1142 . I have tracked that problem to a bug in Navigator when it deals with keep-alive connections. I have been unable to produce a simple test example, or to narrow down which versions cause it. It also appears that there is no workaround other than to disable all keep-alive, something which is not a workaround at all in my books. So I've given up researching this problem for now. Dean From: Marc Slemko To: Apache bugs database Cc: Subject: RE: mod_cgi/1292: post method randomly works (fwd) Date: Wed, 22 Oct 1997 09:14:33 -0600 (MDT) ---------- Forwarded message ---------- Date: Wed, 22 Oct 1997 12:16:45 +0100 From: Lee Pollington To: apache-bugdb@apache.org, dgaudet@apache.org Cc: Jon Marsden , Chris Livermore Subject: RE: mod_cgi/1292: post method randomly works thanks for the response - I tried turning keepalive off, however the problem persisted, we saw it on Netscape 3.03 Unix, 3.01 NT, and 3.0 95, and communicator professional. I didn't test IE. I found no randomness, it always occured, however the data did sometimes get through from the post when the browser was reloaded. I'm sorry to hear you will not be continuing with this, I have had to revert to 1.2.0 (where the problem doesn't exist), and loose the benefits I found in 1.2.4, namely the fewest server errors we have ever had If you want us to liase on a tcp test, or whatever I will talk to our Unix adminstrators about setting something up thanks again for the prompt response Lee > ---------- > From: dgaudet@hyperreal.org[SMTP:dgaudet@hyperreal.org] > Sent: Wednesday, October 22, 1997 1:38 AM > To: apache-bugdb@apache.org; dgaudet@apache.org; > webmaster@cursci.co.uk > Subject: Re: mod_cgi/1292: post method randomly works > > Synopsis: post method randomly works > > State-Changed-From-To: open-closed > State-Changed-By: dgaudet > State-Changed-When: Tue Oct 21 17:38:31 PDT 1997 > State-Changed-Why: > This is probably the same problem as that described in > PR#1142 . I have > tracked that problem to a bug in Navigator when it deals > with keep-alive connections. I have been unable to produce > a simple test example, or to narrow down which versions cause > it. It also appears that there is no workaround other than > to disable all keep-alive, something which is not a workaround > at all in my books. So I've given up researching this > problem for now. > > Dean > From: Dean Gaudet To: Lee Pollington Cc: apbugs@apache.org, Jon Marsden , Chris Livermore Subject: RE: mod_cgi/1292: post method randomly works Date: Wed, 22 Oct 1997 10:28:47 -0700 (PDT) If you can supply us with a full packet tcpdump of the entire session between the client and server then we can study it and decide if it's a browser bug, and if so figure out if we can work around it. If turning off keepalive didn't help you then this is likely a different bug. (Although it gets tiring working around client bugs ...) Do this: tcpdump -s 1514 -w dumpfile tcp port 80 and host client.ip.addr on the webserver as root, and then launch navigator on the client, and cause the bug to occur. Then quit navigator on the client. Then hit ^C on the tcpdump session. gzip the dumpfile and if it's small (<100k) then mail it to us. If it's large then put it somewhere and send us the URL. Note that passwords and such sent from the client will be captured in this, so use non-sensitive data. Dean On Wed, 22 Oct 1997, Lee Pollington wrote: > thanks for the response - > > I tried turning keepalive off, however the > problem persisted, we saw it on Netscape > 3.03 Unix, 3.01 NT, and 3.0 95, and > communicator professional. I didn't > test IE. I found no randomness, it always > occured, however the data did sometimes get > through from the post when the browser was > reloaded. > > I'm sorry to hear you will not be continuing with > this, I have had to revert to 1.2.0 (where the problem > doesn't exist), and loose the benefits I found in 1.2.4, > namely the fewest server errors we have ever had > > If you want us to liase on a tcp test, or whatever > I will talk to our Unix adminstrators about setting > something up > > thanks again for the prompt response > > Lee > > > ---------- > > From: dgaudet@hyperreal.org[SMTP:dgaudet@hyperreal.org] > > Sent: Wednesday, October 22, 1997 1:38 AM > > To: apache-bugdb@apache.org; dgaudet@apache.org; > > webmaster@cursci.co.uk > > Subject: Re: mod_cgi/1292: post method randomly works > > > > Synopsis: post method randomly works > > > > State-Changed-From-To: open-closed > > State-Changed-By: dgaudet > > State-Changed-When: Tue Oct 21 17:38:31 PDT 1997 > > State-Changed-Why: > > This is probably the same problem as that described in > > PR#1142 . I have > > tracked that problem to a bug in Navigator when it deals > > with keep-alive connections. I have been unable to produce > > a simple test example, or to narrow down which versions cause > > it. It also appears that there is no workaround other than > > to disable all keep-alive, something which is not a workaround > > at all in my books. So I've given up researching this > > problem for now. > > > > Dean > > > >Unformatted: