118219 - clicking Reload or Shift+Reload once GETs twice

Status: RESOLVED FIXED

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

Description

[Daniel Ruoso]

The url http://tools.hpg.com.br/cgi-bin/Count.cgi?ft=0&dd=I&df=dbinfovol.dat is
a visit counter.

*** HOW TO REPRODUCE ***
- load this url
- click shift+reload several times
- the number increments 2 every click
- click on the location field and press enter
- the number increments 1
***

I have selected Networking:Cache, because I think Mozilla is making one GET to
discover if the version in the cache is older then the version in the server,
and when it discover that it isn't it makes another GET.

Comment 1

[Daniel Ruoso]

By the way, clicking just reload (without ****) causes the error also.

Comment 2

[Daniel Ruoso]

> By the way, clicking just reload (without ****) causes the error also.
-----------------------------------------------^
sorry... without shiFt

Comment 3

[Chris Casciano]

(NOTE: Tried setting Platform>ALL/OS>ALL, but I don't seem to have permissions,
although I am quite sure I used to be able to to this, but perhaps I've only
done it for my own bugs)

Seeing this behavior on 0.9.9 / Mac OS X also.

Didn't start examining http calls or anything, but a quick tail of my access
logs of a static gif on my own server also showed 2 requests made, but only in
the shift-reload scenario:

127.0.0.1 - - [05/Apr/2002:18:50:24 -0500] "GET /temp/ayard.gif HTTP/1.1" 200 183396
127.0.0.1 - - [05/Apr/2002:18:50:55 -0500] "GET /temp/ayard.gif HTTP/1.1" 304 -
127.0.0.1 - - [05/Apr/2002:18:51:31 -0500] "GET /temp/ayard.gif HTTP/1.1" 200 98304
127.0.0.1 - - [05/Apr/2002:18:51:31 -0500] "GET /temp/ayard.gif HTTP/1.1" 200 183396

The first request is the initial request made.
The second (304) is the (appropriate) response to the normal (non-shift) reload
scenario.
The last two requests came in when I hit shift-reload. Notice the incorrect size
of the first. Of these two requests.
 
Focus into the location bar and simply typing enter produced NO REQUEST for this
test.

NOTE: This was also confirmed via another party in a post to n.p.m.general
earlier today w/ subject: "NS6 requests JPG twice? ?? ???"

Comment 4

[kurros (Wade Majors)]

I'm seeing this with Win32/2002040203 as well.

Confirming.

Comment 5

[Andrew Hagen]

All/all. Summary change for clarity.

Comment 6

[Guillaume Filion]

Does this bug blocks bug 121084?

Does this only occur when "Compare the page in cache to the page on the 
network" is set to "Every time I view the page"? If so, then it's quite 
possible that this bug is causing the behavior of bug 121084.

Comment 7

[Jason Priebe]

I saw this behavior with Mozilla 1.1b under Win2K (also under Linux, not sure of
Moz version).

I watched my server logs, and I noticed that most of the time, both requests
would come in with empty referers.  Occasionally (maybe once every 10-12 pairs
of requests), the second request would have the first request's URL as the
referer (of course, the first and second requests had the same URL, so the
second request had its own URL as its referer).

Some sort of timing issue involved here?

Comment 8

[Darin Fisher]

WORKSFORME linux build 20021023 trunk.

Comment 9

[Jason Priebe]

I thought I was seeing this problem in the 2002110604 build.  It turns out that
the server in question was not sending a Content-Type: header.  So the response
looked like this:

HTTP/1.1 200 OK
Date: Wed, 06 Nov 2002 18:25:31 GMT
Connection: close
Content-Type:

GIF89...<rest of binary deleted for post>

I suspect that the missing Content-Type has something to do with why Mozilla is
reloading the URL.  Once I added image/gif to the Content-Type header, the
behavior went away.  I can't say for certain whether this was the case back in
August when I first saw this behavior.

Comment 10

[Chris Casciano]

just doing some long overdue follow up... WFM 1.3b OS X

Comment 11

[Jim Dunn]

added myself... trying to verify if this is indeed fixed with or without
the patch for bug 121084 and/or bug 83774, so far it looks like it is 
fixed.

Comment 12

[Jim Dunn]

With 1.4a, I am actually seeing a slight regression.  If you click
shift reload OR hit enter in the location field, I am seeing a 
double GET.  However, testing with my fix for bug 89419 fixes
the problems completely.  Gordon, I think we can dup this of
bug 89419 (or you can assign this to me) and when we get that
fix in we can verify it for sure.

Comment 13

[Jim Dunn]

Correction, I am seeing this fixed with:
Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4a) Gecko/20030403
(today's daily build)

Comment 14

[gordon]

Marking FIXED as per recent comments.  Thanks to Jim Dunn and Boris Zbarsky for
recent ImgLib fixes.