View previous topic :: View next topic |
Author |
Message |
Tanner
Joined: 06 Jul 2002 Posts: 55 Location: Germany
|
Posted: Tue Nov 12, 2002 2:33 pm Post subject: Never ending post-loop?!?! |
|
|
If I post something and publish it, the browser´s working and working and...
Then that post was x times published and I have to delete every double post. |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Tue Nov 12, 2002 3:55 pm Post subject: |
|
|
Hi,
A little more detail might prove helpfull.
What version of b2
Did this just start
What changes have you made
Is it always the same number of extra posts
michael e |
|
Back to top |
|
 |
Tanner
Joined: 06 Jul 2002 Posts: 55 Location: Germany
|
Posted: Tue Nov 12, 2002 4:12 pm Post subject: |
|
|
Okay, well...
Now Im using 0.6.1 - but that "bug" occoured in 6.0 too.
Normally, after pressing the submit button, the page ist submitting that post by connecting to the server - see statusbar in browser.
After that the page reloades.
Here it doesn´t stop "connecting" - the longest tim I let it gone was 2 min for a 200 types text!
The I pressed "stop loading page" at my browser - went to my blog:
Wow , 20 times the same post!
Everytime tha error occours! The longer I wait canceling the connection, the more post are there. |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Tue Nov 12, 2002 4:59 pm Post subject: |
|
|
Hi Tanner,
Thanks for the info.
This is a very strange problem, but I don't think this is an issue with b2 directly.
I gather that the problem does not occur every time you post, correct?
Also what browser are you using?
If needed, would you mind if I registered and attempted to post?
michael e |
|
Back to top |
|
 |
Tanner
Joined: 06 Jul 2002 Posts: 55 Location: Germany
|
|
Back to top |
|
 |
terrapin
Joined: 25 Jan 2002 Posts: 3 Location: Washington, DC
|
Posted: Tue Nov 12, 2002 6:17 pm Post subject: Mozilla |
|
|
I just noticed this "bug" within the last few days since I switched to Mozilla. The post screen just hangs even though the material is posted to the site. I eventually just click on "post/edit" to stop the hang.
Another Mozilla-specific issue I have noticed is that when a user posts a comment the page they see *after* posting is the comments page, but it doesn't show their most recent comment. One needs to refresh the page to see the comment. This *could* lead commenters to assume their post wasn't accepted and try and comment again thereby making a duplicate entry.
cheers
-scully
http://flatulent.net |
|
Back to top |
|
 |
Mister44

Joined: 31 Oct 2002 Posts: 237 Location: Philadelphia, PA, USA
|
Posted: Tue Nov 12, 2002 6:24 pm Post subject: |
|
|
I've seen cases where Mozilla 1.1/Win32 will repeat an operation several times. Normally it seems to happen HTTP1.1 Pipelining is enabled and the server in question is responding slowly.
I'm looking thru bugzilla to see if its documented in there. (But bugzilla is a serious PITA, so if anyone else can confirm that'd be great.) |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Tue Nov 12, 2002 7:19 pm Post subject: |
|
|
Hi,
Follow up. After registering on Tanners site and posting a simple test post using IE 5.5 on a mac, the same symptom occurred. However, the most that got posted was 2. In some cases the post timed out.
Tanner. For test purposes, turn off trackback and pingback in your config and attempt a post. Does the issue persist?
Also please confirm if you have made changes to the b2edit.php or the b2edit.form.php files.
Michael e |
|
Back to top |
|
 |
Mister44

Joined: 31 Oct 2002 Posts: 237 Location: Philadelphia, PA, USA
|
Posted: Tue Nov 12, 2002 7:49 pm Post subject: |
|
|
Hehehe - just managed to have it happen on my journal. Of course, my b2 install is modded all to hell so it may not be useful in debugging other than to say that its not an isolated occurence.
ok - so here is some info about it - I was tracking back to one of my own entries and pingback was turned on. It was stuck for about 3 minutes before i killed it (hit stop in the browser) and it posted twice. it did -not- register the trackback. I'm using Moz1.1 on WinXP Pro. |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Tue Nov 12, 2002 10:18 pm Post subject: |
|
|
Hi again.
Are we saying that a post could be re-issued under some conditions that cause b2 to enter the data twice?
If true, does that also say that the browser has a role in this issue by attempting an "auto" retry of the post/request?
Mister44 are you aware of a interface spec. (http or other) that would require an auto retry? It doesn't make sense to me to do that, but I am weak in the details re this protocol spec.
The flow...
When the post is made, b2edit is going to cause:
include b2header - just the includes no response to client
update the db
rssupdate - Don't think anything goes back to client
pingWeblogs - Don't think anything goes back to client
pingCafelog - Don't think anything goes back to client
pingBlogs - Don't think anything goes back to client
maybe post_pingback
maybe trackback_urls
Then send header to client to access b2edit.php - first response to client?
It would appear that maybe the ping logic has an issue with slow or unresponsive servers....
Michael e |
|
Back to top |
|
 |
Mister44

Joined: 31 Oct 2002 Posts: 237 Location: Philadelphia, PA, USA
|
Posted: Tue Nov 12, 2002 10:48 pm Post subject: |
|
|
macshack wrote: | Are we saying that a post could be re-issued under some conditions that cause b2 to enter the data twice?
If true, does that also say that the browser has a role in this issue by attempting an "auto" retry of the post/request?
|
It certainly does appear that way. My access log shows Code: | 68.45.252.252 - - [12/Nov/2002:13:45:53 -0500] "POST /journal/b2edit.php HTTP/1.1" 200 - "http://frcooper.com/journal/b2edit.php" "Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.1) Gecko/20020826"
68.45.252.252 - - [12/Nov/2002:13:47:55 -0500] "POST /journal/b2edit.php HTTP/1.1" 200 - "http://frcooper.com/journal/b2edit.php" "Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.1) Gecko/20020826" |
macshack wrote: | Mister44 are you aware of a interface spec. (http or other) that would require an auto retry? It doesn't make sense to me to do that, but I am weak in the details re this protocol spec.
|
I haven't seen anything in the specs that would indicate an auto-retry. I'm currently leaning towards the idea that the header directives b2edit.php issues are getting screwed up. Possibly issuing a refresh? We've seen the behaviour with both IE and Moz. I'll try to reproduce it in Opera. Also if there's access to a logging HTTP proxy, we could get a record of the traffic. I don't access to one at the moment. Although I suppose I might be able to get one up in a couple hours. Another thing to try would be SSL - Anyone got b2 under https?
macshack wrote: | It would appear that maybe the ping logic has an issue with slow or unresponsive servers....
|
Well, as it happened when I ping/tracking back to my own server, neither of those cases should be true. I don't see that there were any zombied processes for apache or mysql at the time.
I'm curious. Can you ping/track back to my blog? Just pick a random entry. Maybe its something in the server configuration for keepalive, or connection timeout? |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Wed Nov 13, 2002 12:58 am Post subject: |
|
|
Hi Mister44,
I enabled ping/tracbacks on my internal server and did a post with a link to your site post #326.
The post took a little longer, but no duplicates and your site recorded the ping.
(This is a generic b2 v 0.6.1 no mods)
Remember the other thread where there was discussion on the new header Michel added to blog.header.php? I wonder if that may be tied to this issue:
Code: |
/* Sending HTTP headers */
$last_modified_header = mysql2date('D, d M Y H:i:s', get_lastpostdate());
@header ("X-Pingback: $pathserver/xmlrpc.php");
@header ("Last-Modified: $last_modified_header");
@header ('ETag: "'.md5($last_modified_header.$pagenow).'"');
|
michael e |
|
Back to top |
|
 |
macshack
Joined: 17 Jul 2002 Posts: 1204 Location: Phoenix, Az
|
Posted: Wed Nov 13, 2002 1:02 am Post subject: |
|
|
Hi again Mister44 & Tanner,
Just a reminder, my post that I did on Tanner's site was just text, no urls. However, Tanner has ping/trackback enabled.
Tanner, have you done a test post with ping/trackback disabled? And do you get the same symptom?
Michael e |
|
Back to top |
|
 |
Mister44

Joined: 31 Oct 2002 Posts: 237 Location: Philadelphia, PA, USA
|
Posted: Wed Nov 13, 2002 1:27 am Post subject: ETag |
|
|
Reading over the RFC for HTTP1.1, I can't see how it would be even involved unless the browsers are using it for cache management... and they shouldn't be caching form submittals. |
|
Back to top |
|
 |
mikelittle
Joined: 11 May 2002 Posts: 376 Location: UK
|
Posted: Wed Nov 13, 2002 1:43 am Post subject: |
|
|
macshack,
that last modified header will certainly explain why recent comments are not seen without a forced refresh. By definition the comment must be a later date than the latest post, but the headers says last modified is the date of the post.
The browser sends a HEAD request and gets the same date back, so it can use the cached version. Similar problem with proxy servers.
I disabled it on my site and that problem disappeared.
I've not however seen the 'looping' problem before or after removing the header.
One point, b2edit.php does send a 302 location: header after posting. But this shouldn't make the browser re-submit the form.
Mike |
|
Back to top |
|
 |
|