Old Proxomitron Forums
Proxomitron Program - discussions welcome => Questions and Answers => Topic started by: MorpheusDreamlord on June 17, 2002, 05:01:55 PM
-
I've noticed an error message in the log window which occurs on (before?) each 1/2 page failure..
** 2511 Socket Error 10038 for request flush **
'2511' obviously changes. Is this what's causing the 1/2 page load problem?... Anyone else seen this error
<Question/Problem: Socket operation on non-socket.
Answer/Solution: An operation was attempted on something that is not a socket. The specified socket parameter refers to a file, not a socket.
...Found this tho, maybe the problem is linked to local files?? (damn, I'm really scratching here.. LOL>
|
Come to the Dreaming...
Edited by - MorpheusDreamlord on 17 Jun 2002 18:45:01
-
What request/reply does the connection number tie to? Also exactly what filters are you using at the time and are you doing any RDIR's?
-
I should also ask if anyone else is getting the same error. Actually I'm not sure everyone is reporting the same problem, since you seem to also have crash problems the others don't. It's possible something is corrupting Proxomitron's memory, which could cause both types of errors.
Still, I need to know which URL the error ties to - the stylesheet or the page itself. No matter what I try I just can't get this to happen on my system - at least not with any of the filter sets I've tried.
-
I've also gotten that error. I haven't seen it with the half loaded problem tho.
To test it, I went back to using <start> to insert css. At first I couldn't get the bug to reappear, but after I started editing my config & loading a crap load of pages, it came back.
This time I kept the log window open, but..
** 2511 Socket Error 10038 for request flush **
did not appear when the half loaded pages occured.
Although, I found by unchecking "Local connections persist for xx seconds" I was able to load the page without problems.
The problem was at Y!Groups.
~Just another blind stab in the dark. JD
--------
"Imagination is more important than knowledge" - Einstein
-
quote:
I've also gotten that error. I haven't seen it with the half loaded problem tho.
Where have you seen the error then and are you sure it was that exact error (and not a similar one)?
Also what timeout did you have on the local connections and how many connections do you have your browser set to make?
-
At random places.. I just had one now.
+++GET 5200+++
GET /pforum/forum.asp?FORUM_ID=3 HTTP/1.1
User-Agent: Opera/6.03 (Windows 98; U) [en]
Host: asp.flaaten.dk
Accept: text/html, image/png, image/jpeg, image/gif, image/x-xbitmap, */*
Accept-Language: en
Accept-Charset: windows-1252;q=1.0, utf-8;q=1.0, utf-16;q=1.0, iso-8859-1;q=0.6, *;q=0.1
Accept-Encoding: gzip, deflate
Referer: http://asp.flaaten.dk/pforum/
Cookie: <snip>
Cookie2: $Version="1"
TE: deflate, gzip, chunked, identity, trailers
Connection: keep-alive
+++CLOSE 5198+++
Client closed: total 3
Client closed: total 2
Connection Reused: 5198->5200
** 5200 Socket Error 10038 for request flush **
Connection Reused: 5182->5200
BlockList 5200: in CookieList, line 23
I have my timeout at 30.
My connections are at,
8 per server
12 Max
Although, I haven't seen Proxo show more then 8 active connections.
~JD
--------
"Imagination is more important than knowledge" - Einstein
-
I have those socket errors frequently too, although I do not have problems with images. Naoko 4.3 here.
-
Just to let everyone know I've been looking into the error and I think I found the cause (of the error at least). It's an attempt to reused a stale socket handle, but I'm not sure if it could be related to the half-loading bug or not.
In most cases the error would only cause Proxomitron to retry on a fresh connection, and this happens before the page begins to load (when the server is first contacted). Since you never know when a remote server will time out a connection, the program actually expects some sort of error at that point, but servers normally don't time-out actively downloading connections.
Anyway, here's something to try...
In = FALSE
Out = TRUE
Key = "If-Modified-Since: Always reload pages (Out)"
URL = "[^/]++local.ptron/"
Match = "*"
This (in a round-about way) should avoid the error causing the stale handle to be used, and might at least confirm/deny if it's the problem.
Thanks,
Scott
Edited by - Scott Lemmon on 19 Jun 2002 02:32:51
-
Hi Scott,
for me all problems disappeared after adding local.ptron to my bypass list.
Changing filters back to dhtml instead of local.ptron/ also works (but this doesn't work over a LAN).
I don't see why the 'site' local.ptron would have to be filtered in the first place, there can be no possible security risk attached, right?
-
You should be able to bypass it without problems. In fact I had considered just making it always bypassed, but there's the posibility someone may want to filter it for some reason - perhaps to add some dynamic element to the page.
-
quote:
Anyway, here's something to try...
(filter)
This (in a round-about way) should avoid the error causing the stale handle to be used, and might at least confirm/deny if it's the problem.
Hi Scott, I just installed the last Opera bèta (6.04 b1129), and the half-loading pages reappeared. I have been following this topic with great interest of course, so tried the filter you posted: using the above filter does not prevent incomplete page loads, nor does the socket error message disappear. (The relation between this socket error message and incomplete loading pages is not as easy as it seems: the socket error message does not always appear with partially loaded pages, but if it appears, the page will not load completely).
The original page request still occasionally stops loading once prox.css is called -- but only if prox.css is included on <start>. Rewriting that filter according to altosax' suggestion (http://"topic.asp?TOPIC_ID=604&whichpage=6&ARCHIVE=#2341") completely eliminates the partial loads.
Edited by - Jor on 19 Jun 2002 15:11:34
-
quote:
The relation between this socket error message and incomplete loading pages is not as easy as it seems: the socket error message does not always appear with partially loaded pages, but if it appears, the page will not load completely
This is not my experience - I never get half-loaded pages (using 4.3 and opera 6.03), but I do occasionally get the socket errors - but to no ill effect seemingly.
-------------------------
|David Gallagher
|dave at smokeajay.co.uk
-------------------------
-
quote:
This (in a round-about way) should avoid the error causing the stale handle to be used, and might at least confirm/deny if it's the problem.
Hi Scott, I just installed the last Opera bèta (6.04 b1129), and the half-loading pages reappeared. I have been following this topic with great interest of course, so tried the filter you posted: using the above filter does not prevent incomplete page loads, nor does the socket error message disappear.
[/quote]
You might want to check that the filter is working properly. If so you should never see a "Not modified" reply when Proxomitron is asked for a file from http://local.ptron/. Check to make sure the If-Modified-Since: header is really being removed for local requests.
-
Ack... my mistake.
I had local.ptron in my bypass list, so this filter of course never matched.
It's working now, and indeed, no more partially loaded pages or socket errors.
-
Now that WAS fun!!
I did this:
In = FALSE
Out = TRUE
Key = "If-Modified-Since: Always reload pages (Out)"
URL = "[^/]++local.ptron/"
Match = "*"
(I edit the .cfg, then save it and open it in proxo)...
BOOOOM!, wsock.vxd bsod, various other wsock(xxx).xxx errors (after pressing a key), then vmm.vxd died, and bingo!! auto reboot!!.
LOL!!
|
Come to the Dreaming...
Edited by - MorpheusDreamlord on 20 Jun 2002 16:45:14
-
I think you need to get a new PC

Seriously, something sounds very wrong at the driver or hardware level. You're not over-clocking or have your memory latency tweaked too high by any chance?
Other things high on the list of potential problem causers are firewalls and virus-scanners. Try disabling anything of that sort and see if the problems persist.
It might help to post a list of all (and I do mean all) running processes on your system. Also please report the *full* error you see - it'll usually include the driver's name.
-
No, basically everything is standard, mem latency is lower of two options in bios (7/9 and ras/cas 3), no overclock p3-800, no firewall, NAV virus (but i don't use the monitor, I do manual scans)
Basically at the moment its:
poppy (email monitor)
cleansweep (+ csinject and monwow, it's two attendant progs)
IE6 (about 5-8 windows at once)
Proxo
an explorer window open to my downloads folder
Massdownloader 1.2 (like getright, but allows number order downloads
I'm using ME installed via 98Litepro 4.6 (custom installer removes bumf like medplayer 7.1 -> 6.4 swap)
and thats it. I've just totally killed out dialup network and all networking in control panel, reboot, and reinstall dialup network. No lan, no nuttin (LOL).
So far, the reinstall may have helped (not tried that filter yet tho, not game to...).
According to WinUpdates I needed the slow shutdown patch for large cache hard drives, and it seems they lied and the latest IE patch does NOT include the older ones as well.. it downloaded and installed the older ones (just looked - it seems it may have overwritten the newest patch, the Q number is different.. damn stoopid MS!!)
<edit> I've just had this:
Connection Reused: 2413->2421
** 2421 Socket Error 10038 for request flush **
Connection Reused: 2414->2421
I notice that the socket 2413 was tried for 2421, failed and 2414 was used instead (am I reading that right?) - maybe it is associated with the local.ptron files and reuse of socket for online files? Or am I TAMA again?? <>
Edited by - MorpheusDreamlord on 21 Jun 2002 18:03:52
-
quote:
I'm using ME installed via 98Litepro 4.6 (custom installer removes bumf like medplayer 7.1 -> 6.4 swap)
Stuff like this can sometimes cause problem sometimes if it leaves out files other programs may expect to be there (especially other MS programs). It may also confuse windows update. The IDE patch might be a good idea - the one BSOD you mentioned before was in a file system driver.
quote:
Connection Reused: 2413->2421
** 2421 Socket Error 10038 for request flush **
Connection Reused: 2414->2421
I notice that the socket 2413 was tried for 2421, failed and 2414 was used instead (am I reading that right?) - maybe it is associated with the local.ptron files and reuse of socket for online files? Or am I TAMA again?? <>
That's normally how it works. They're just connection numbers the program uses, (so don't have any deeper meaning), but whenever it tries to re-use one it expects it might fail. It will then retry on a different connection which may be another re-used one or a new one depending on what's available.
I'm guessing you don't have the Last-Modified filter active right? It should stop the error.
-
Yas, 98Lite kills WinUpdates, had to activate that manually (but it's ok now).
But it has been well tested by the guy, It's not just a quickee hack (been around since 98 v1, and he's working on an XP installer (wowee, kill cd writing and movie maker, before the install!!).
I noticed my earlier erors were filesystem related too.
Re- the connection reused bit, I was linking the first reuse try with the error about request flush, then another reuse come in - it seems it trys the reuse, but can't clear it (again, I'm probably TAMA again - I do that a lot too, hehehe)??
No i don't have Last-Modified filter active - that's what 'activated' my lasy wsock.vxd bsod, not game to retry him.
Scott, thanks for the interest. I accept it's not Proxo. But I do think it's Proxo opening up the can-o-worms of something already here, maybe. A bit like a stress test..
<crossfinger>(BTW - all is fine so far after the dialup reinstall..)
</crossfingers>
<edit - just found this:
Changes.txt:
Naoko 4.2.2 (Beta Test version, please do not re-distribute yet)
* Added persistent connection support to local connections as well
as remote. For best results make sure your browser is set to use
HTTP/1.1 through the proxy and has pipelining enabled (if it's an
option). Also I recommend about 6-8 simultaneous connections. With
this, Proxomitron now supports nearly all of HTTP/1.1's optional
features.
I DON'T have 'HTTP/1.1 through the proxy' enabled, just the HTTP/1.1 setting in IE. Does this alter anything vis Proxo?
Edited by - MorpheusDreamlord on 21 Jun 2002 20:06:00
-
Here is what i observed (XP Pro, IE6, Kerio FW, Naoko 4.3).
For about one week i had these filters active for all sites:
If-Modified-Since (out)
If-None-Match (out)
Last-Modified (in)
I didn't get any halfload (nor a BSOD),
though i had to reload pages more often then with 4.2 to see anything at all,
but that might not be related.
Then i turned these filters off, and got some halfloads on this forum.
After half a day i turned on your workaround filter and still got them, maybe less often, not sure.
Afer an hour, while reloading a forum page, the firewall driver blue screen came back.
I didn't get it managed to have the log window open during a halfload,
so i don't know about any socket errors.
Both, local.ptron and asp.flaaten.dk/pforum/ are on a bypass list for all web filters
(URL header filter with $FILTER(False)).
edit: persistant connections were set at 60 sec.
sidki
Edited by - sidki3003 on 23 Jun 2002 13:10:19
-
quote:
Both, local.ptron and asp.flaaten.dk/pforum/ are on a bypass list for all web filters
If local.ptron is in your bypass list, then the header filter I posted will never be called.
It's job is to filter if-modified-since headers from local.ptron requests - you'll likely need that header for other URLs.
-Scott-
-
It's not on the general bypass list.
this is the filter:
[HTTP headers]
In = FALSE
Out = TRUE
Key = "URL: Bypass selected (Out)"
URL = "$LST(Bypass_Web)"
Replace = "$FILTER(False)"
I thought it would just bypass the web (vs. header) filters.
-
Should be easy enough to see in the log. When local.ptron is called you should never see an if-modified-since header and the reply should always be a "200" code.
-
That's what's happening
. Anyway, just thought the report might be of use.