85127 - PHP output cached

Status: VERIFIED DUPLICATE of bug 79983

(Core :: Networking: HTTP, defect)

Description

[Trever Adams]

I can get out put from any php script that uses post instead of get.  Problem
is, (shift, ctrl, alt, etc) reload and things like that don't refresh the page.
 I have to go back several layers and try again.  This is very annoying and
doesn't happen ins NS 4.xx.

Comment 1

[gordon]

Darin, how can we be reusing old data for a POST transaction?

Comment 2

[benc]

Are you seeing any network activity, unless the engineers know what you are
talking about implicity, you need to provide a more detailed description.

Comment 3

[Trever Adams]

[e-mail] it isn't that you are using old data, it seems that it just
rejects any new output from PHP scripts, if it even asks.  Checking in answer to
B. Chuang.  No, normal reload doesn't send any network activity, on a cgi output
it should.  However, it seems the newer Mozilla browsers are sending data on
shift, ctrl, alt, etc. However, the newer output doesn't always replace what
Mozilla has.  Cache bug maybe?

Comment 4

[basic]

reporter could you provide a testcase?

Comment 5

[Trever Adams]

phpnuke.org used to be one.  The only one I know of still (as in that I still
use) is an inhouse app that is behind a firewall.  To be honest, I think this
bug may be fixed for all sites that have a domain name.  I can only reproduce it
with internal sites (which have no DNS name).  Control-Reload still produces no
network traffic in such cases, though it does on phphuke.org.

Is there a way I can get the output of the page or some such to you as the
example since you can't use the internal site (though I may be able to hand off
the source code, though it wouldn't do much good without the database and I have
no example database and I can't share the live one.

Comment 6

[benc]

This bug needs more data. Can you provide a packet trace and a specific cache
setting that has a problem?

Comment 7

[basic]

reporter: sorry not to have asked this before, which build are you using?

Comment 8

[Trever Adams]

I may or may not have time providing the tcpdump and such today.  However, cache
setting is 'Automatically'.  However, The last test I did was either with
2001070608 or 20010704xx. I don't remember, but both still show the problem I
described with the last report.

Comment 9

[Trever Adams]

Sorry, Linux and only Linux.  Though it may affect other OSes as well.

Comment 10

[basic]

Reporter please do a packet capture

Comment 11

[Trever Adams]

Ok, sorry I have been extremely busy. Doing it right now. I am not capturing the
data of the packets, just the packet trace. If you need more than a normal
tcpdump, please let me know the switches. I will be incommunicado on July 18-20
because of work and family events, so if I am needed, try to get me before then.

Here it is.

Upon looking at the log, it appears it is now asking the proxy to show it the
information.  I has never done this before (yes, I have done previous traces). 
Maybe it is a squid or mozilla<=>squid bug?  Before the mail connection I was
just doing normal reload, that is the one that never used to work.  The ones
after are normal, shift, and control reload.

Comment 12

[Trever Adams]

Attachment: Packet trace of behavior

Comment 13

[Christopher Aillon (sabbatical, not receiving bugmail)]

Bug 91254 has a reproducable testcase.

Comment 14

[Keyser Sose]

*** This bug has been marked as a duplicate of 79983 ***

Comment 15

[John Unruh]

Verified dupe.