Post by: Scott Lemmon on June 10, 2002, 02:32:16 AM
Well it's finally here. Naoko 4.3 is available at...
http://proxomitron.cjb.net/
Enjoy!
-Scott-
Post by: hpguru on June 10, 2002, 03:12:27 AM
Post by: sidki3003 on June 10, 2002, 03:19:18 AM
Post by: dave1006 on June 10, 2002, 03:32:24 AM
Just installed WinXP Pro, it's 3:30am and im about to fall asleep sitting here, and now i get a new Proxo version to play with too! No sleep for me tonight then :)
-------------------------
|David Gallagher
|dave at smokeajay.co.uk
-------------------------
Post by: JD5000 on June 10, 2002, 04:13:43 AM
--------
"Imagination is more important than knowledge" - Einstein
Post by: Arne on June 10, 2002, 06:49:05 AM
Best wishes
Arne
Imici username: Arne
Post by: Jor on June 10, 2002, 12:49:38 PM
Filter question: $DTM() only works in replace, so if I want to apply this to header filters the follwing construct has to be used:
In = TRUEThis will return an empty ETag header string, is there a way to rewrite this filter to prevent this?
Out = FALSE
Key = "ETag: Always cache (in)"
Match = "(?*) "
Replace = "$LOG(C$DTM(c: )Header ETag removed: )"
(If $DTM() could be modified to work in Match, this would also do the trick).
Post by: Jor on June 10, 2002, 02:35:35 PM
Testing if it is my config, or the new version...
Post by: sidki3003 on June 10, 2002, 02:52:50 PM
edit: I know that might well be a bug in the driver.
Edited by - sidki3003 on 10 Jun 2002 16:10:25
Post by: Scott Lemmon on June 10, 2002, 03:11:44 PM
4.2.2 should not be used at this point. If it works it's only out of luck.
Edited by - Scott Lemmon on 10 Jun 2002 16:27:57
Post by: Arne on June 10, 2002, 03:39:02 PM
Best wishes
Arne
Imici username: Arne
Post by: Jor on June 10, 2002, 05:01:32 PM
Crashes occured both with Mozilla 1.0 Gold, and Opera 6.04.
My firewall is Kerio Personal Firewall, version 2.1.4 (driver 3.0.0)
The actual crash is in fwdrv.sys, the error is DRIVER IRQ NOT LESS OR EQUAL.
The not completely loaded pages can't be easily reproduced, but it appears to only happen on pages that use large tables (such as this forum).
Post by: sidki3003 on June 10, 2002, 07:42:51 PM
quote:
My firewall is Kerio Personal Firewall, version 2.1.4 (driver 3.0.0)
The actual crash is in fwdrv.sys, the error is DRIVER IRQ NOT LESS OR EQUAL.
Same here (XP IE6).
I wonder whether hpguru gets it running.
I think he uses KPF 2.1.4 as well.
Post by: hpguru on June 10, 2002, 11:06:02 PM
I haven't had any problems at all with 4.3. It's working great here.
Facing each other,
a thousand miles apart.
Post by: Scott Lemmon on June 10, 2002, 11:13:44 PM
quote:
The problem appears to be with the $LOG() command, disabling it completely seems to keep the crashes away.
Odd that - can you give me some examples of how you were using the command and the type of data you were sending to it? Thanks!
-Scott-
Post by: sidki3003 on June 11, 2002, 12:48:43 AM
After removing the $LOG commands, prox 4.3 is running 7 hours straight (fingers crossed
).I used it only with header filters like so:
Key = "ETag: Kill (always cache) (LOG) (In)"
Match = " $LOG(CHeader ETag killed: )"
Key = "Content-Type: Fix MIME types (mod S3) (In)"
Match = "(text/*|^?|unknown)1 $URL([^/]+*.([a-z0-9]+{2,5}(^?)&&$LST(MIME-List)))"
Replace = " $LOG(CHeader Content-Type: 1 to )"
I forgot: I chained a caching proxy between the browser and prox,
but it speaks HTTP 1.1 and doesn't disturb windowsupdate.
Edited by - sidki3003 on 11 Jun 2002 02:11:08
Post by: Scott Lemmon on June 11, 2002, 01:16:23 AM
quote:
I think i can confirm that. After removing the $LOG commands, prox 4.3 is running 7 hours straight (fingers crossed
I used it only with header filters like so:
Key = "ETag: Kill (always cache) (LOG) (In)"
Match = " $LOG(CHeader ETag killed: )"
Hm, might be part of the problem. You should never place a command like LOG directly after a wildcard - otherwise it might result in the $LOG command being called hundreds of times! Consider a match like "*a": it must scan through each character in the string looking for an "a". Likewise "*$LOG()a" will scan through the string looking for "a" but also call LOG at each test! Probably not what you want.
In your example though, since nothing follows LOG, the first test will be "true" and will always be empty. Variables aren't set until that phase of the match is complete, so even though it still shouldn't cause a crash, you'll never see what you intended in the log.
A better way to write it would be...
&$LOG(CHeader ETag killed: )
Since the "&" separates the two commands. First " " is completely matched and set, only then the LOG command is called.
-Scott-
Post by: sidki3003 on June 11, 2002, 01:33:15 AM
Thanks for the help
Post by: Jor on June 11, 2002, 02:21:30 AM
quote:
You should never place a command like LOG directly after a wildcard - otherwise it might result in the $LOG command being called hundreds of times!
Is it possible this caused the crashes? I went over my filterset, and adjusted all cases (added ampersands in between... no crash so far.
Post by: Scott Lemmon on June 11, 2002, 02:37:49 AM
Post by: sidki3003 on June 11, 2002, 03:38:31 AM
Post by: Scott Lemmon on June 11, 2002, 03:50:01 AM
When the window's closed a bunch of the work is bypassed and the LOG command does very little. That might indicate if the problem was really in the command itself or how it was called. I'm trying using a similar filter to see If I can get anything to happen - so far no problems though.
Thanks!
-Scott-
Edited by - Scott Lemmon on 11 Jun 2002 04:50:23
Post by: sidki3003 on June 11, 2002, 03:52:37 AM
quote:
was Proxomitron's log window open or closed when the crashes happened?
In my case it was closed.
Post by: Jor on June 11, 2002, 04:01:14 PM
Pretty sure the crash is due to a bug in Kerio, I have uninstalled it yesterday, and have had no crash yet.
Post by: MorpheusDreamlord on June 11, 2002, 04:54:48 PM
<edit - I've found one thing - some pages are cut off after 1/10 th of the load, a refresh is needed to get them fully....>
<edit -
<snipped out start>
<link rel="chapter forum" href="viewforum.php?f=17" title="3rd HOTU Su
<!--//--><script src="http://Local.ptron/PrxEnd.js"></script><!--//-->
Note the end at "HOTU Su"? This is from the underdogs site BTW..>
Just outta interest, the log command JUST sends notification to the log window, right?... I accept that it's interesting, but why bother - it's getting to be like Linux almost - spending so much time just wowing over the mechanics of it working, and just not leaving it to get on with working. The only time i've had the log window open was just to get an idea of which filters got activated the most, in order to move the to the top of the cfg.
(Hey you guys that love the techy stuff - don'e flame me, just trying to say that the log command seems a littel bit of a waste of time).
(Hey Boss - not complaining about Proxy43 either, she's GREAT!!!)
|
Come to the Dreaming...
Edited by - MorpheusDreamlord on 11 Jun 2002 17:58:02
Edited by - MorpheusDreamlord on 11 Jun 2002 20:03:17
Edited by - MorpheusDreamlord on 11 Jun 2002 20:13:59
Post by: pooms on June 12, 2002, 12:02:40 AM
use it just for Browser to Server HTTP, but also in a more peer-to-peer
environment. In particular, the HTTP Response does not always go back to
a Browser, and because Proxomitron's HTTP Message Log does not have an
option to display the contents of the HTTP Response, I can't see it.
I can turn on View Posted Data to see the data content of the HTTP Request,
but there is no equivalent for the return data. So now I can write little
filters that look for things of interest in the response data and LOG them
to the Message Log. So it's a very practical new feature for me.
(Now if I could only make the content filters selectable as Out or In
filters, like the header filters, it would work even better in a
peer-to-peer situation
)Post by: Jor on June 12, 2002, 01:13:33 AM
I reinstalled Kerio, and almost immediately got a crash in fwdrv.sys again. I have no idea what is the problem here, but it would seem something in Prox 4.3 can cause Kerio to go insane.
I'm using another firewall now (Conseal) without any problems, maybe Kerio's next version will be better.
Post by: Scott Lemmon on June 12, 2002, 01:19:31 AM
quote:
<edit - I've found one thing - some pages are cut off after 1/10 th of the load, a refresh is needed to get them fully....>
It's always possible a busy site (or general network problem) may cause a connection to drop prematurely. I've not seen this kind of behavior, but some people using Kerio firewall were reporting similar things happening (see some of the other posts here). Might be the firewall in that case, but it's hard to say for sure.
It's also possible some filter you're using may be doing this - especially if the filter a) has a very large byte limit b) tries to insert all the original text it matched plus some additional text, and c) has multi-match enabled.
Even though the replace buffers are dynamic, this only means the combined output of multiple filters with big bounds won't exceed the buffer, but it's still possible for a single filter to do so. Still, it's a lot easier to account for that in a single filter than try and guess about multi-filter interactions. Just remember the total output needs to be under 32k for a single call of the filter when using multi-match (so don't match 32k of data and then stuff a 10k script on top of that for instance).
-Scott-
Edited by - Scott Lemmon on 12 Jun 2002 02:30:51
Post by: ridgydidge on June 12, 2002, 06:32:41 AM
doesn't happen in 4.2 using the same filters
Post by: JakBeNymble on June 12, 2002, 08:27:54 AM
No Problems here so far! And it does seem to be faster than the other version that I was Using. "SCOTT" thanks for working on that connection Problem!
And For Making Proxomitron available to us! I know that really proxo didn't have a problem, but other factors involved caused the connections to build up. Well You all have the Best Evening ever!
"Best Wishes",
'jak'
Post by: MorpheusDreamlord on June 12, 2002, 09:09:20 AM
quote:
I'm also getting the page stop problem on all sorts of pages and links opening to images - requiring a refresh in 4.3.
doesn't happen in 4.2 using the same filters
Yup, that's my point exactly, nothing has changed here on my HD except the proxo executable, and we're getting the pages cuts offs..
Update - if you forve a dbug reload od a bad page, turn dbug off then refresh the page, all is fine. (The dbug page shows all filters are working fine - it's almost as if Proxo is scanning the incoming htm TOO fast, and on slow connections assumes no more code is the end of file (??????)). I'm pretty sure is not a filter problem.
|
Come to the Dreaming...
Edited by - MorpheusDreamlord on 12 Jun 2002 10:14:31
Post by: dave1006 on June 12, 2002, 03:21:08 PM
I am, however, not using any of the new commands.
-------------------------
|David Gallagher
|dave at smokeajay.co.uk
-------------------------
Post by: sidki3003 on June 12, 2002, 03:32:23 PM
However, i got a response from Kerio.
They are looking into it
Post by: Bes on June 12, 2002, 05:34:31 PM
quote:
However, i got a response from Kerio.
They are looking into it
That's great to hear, at least, there still are some
companies who DO listen to their customers, and then there's
Scott. There *IS* still hope
Post by: Scott Lemmon on June 12, 2002, 11:03:06 PM
Unfortunately many people are using configs that do all kinds of unusual things, and all I can vouch for is the default set. It's possible some unusual odd syntax in a filter worked (or at least was ignored) in previous version that may cause trouble now.
Post by: ridgydidge on June 13, 2002, 08:29:51 AM
I did as you suggested and loaded the default.cfg that comes with 4.3 - no page complete problem - all pages loaded fully
back to my preferred set JD500.cfg and the problem reoccurs
What a pity, JD's config rocks.
Post by: JD5000 on June 13, 2002, 10:05:05 AM
quote:
The page not loading problem
I did as you suggested and loaded the default.cfg that comes with 4.3 - no page complete problem - all pages loaded fully
back to my preferred set JD500.cfg and the problem reoccurs
What a pity, JD's config rocks.
Yikes! That sucks.
Can you give me some url's to test?--------
"Imagination is more important than knowledge" - Einstein
Post by: MorpheusDreamlord on June 13, 2002, 04:29:09 PM
Sorry, Scott - I'll wait for 4.3.1 heheh
(Quote - 'It's not possible for an application to read data "too fast" or "too slow"' - - yup. I understand that, I was really just trying to describe what it looked like this end.)
Edited by - MorpheusDreamlord on 13 Jun 2002 17:34:48
Post by: Scott Lemmon on June 13, 2002, 05:04:04 PM
As for the half-loading, if someone can find an example of which filter may be causing it that would help. As I've said, I've never seen this happen, but I only use my own filters. One guess, maybe a filter is trying to stuff to much into a multi-match replace in some cases? It might cause the replace output to be truncated.
Post by: Jor on June 13, 2002, 05:11:49 PM
Also... the pages are not half-loading (it is not similar to the problems I had with half loading images in older Opera versions). They seem to break at a random point, for example I had it in a thread on this forum about an hour ago: the pages stopped loading on the 'All Forums' link above. After that, the content which Prox adds on <end> was added. This would seem to indicate Prox thinks the page is completely loaded...
A thought, could this have to do with Pipelining? I noticed it in Opera, Mozilla, and MSIE.
As for the BSOD's: the issue definately is in Kerio (at least on my system). Haven't had a lockup since I deinstalled it.
Edited by - Jor on 13 Jun 2002 18:17:11
Post by: MorpheusDreamlord on June 13, 2002, 05:28:01 PM
http://www.webknacks.com/aptest_c2.htm
Is a popup test site. The page says
'There is a very stubborn popup on this page...
Some programs can get into a loop on this test.'
Sometimes the page hangs for 5 secs showing
'Ther'
then loads.
Sometimes the page loads fully-cut at the 'Ther' bit
A refresh fixes it, but the </html> at the end is there!!!
Doing nothing but testing this at the moment - been testing various filters for 3 hours now!!
Other than that, Jor's right - it seems to be random, NEVER happens when the dbug is on, a refresh will fix (sometime not on 1st refresh tho)
And the Log shows (some bypasses and closes removed for space)(not the flush errors - what's that?)
BlockList 58: in MIME-List, line 28
+++CLOSE 58+++
Match 57: Java redirects to links
<end> 57: Kill event handlers
<end> 57: Java @ End
+++CLOSE 57+++
<start> 59: Java @ Start
<start> 59: Kill window manipulation
<start> 59: Kill java functions
+++CLOSE 60+++
+++CLOSE 61+++
+++CLOSE 62+++
+++CLOSE 63+++
Browser reload detected...
<start> 64: Java @ Start
<start> 64: Kill window manipulation
<start> 64: Kill java functions
<end> 59: Kill event handlers
<end> 59: Java @ End
+++CLOSE 59+++
<end> 64: Kill event handlers
<end> 64: Java @ End
+++CLOSE 64+++
<start> 78: Java @ Start
<start> 78: Kill window manipulation
<start> 78: Kill java functions
<end> 78: Kill event handlers
<end> 78: Java @ End
+++CLOSE 78+++
BlockList 86: in Bypass, line 20
** 86 Socket Error 10038 for request flush **
+++CLOSE 83+++
BlockList 87: in Bypass, line 20
+++CLOSE 91+++
BlockList 92: in Bypass, line 20
** 92 Socket Error 10038 for request flush **
** 92 Socket Error 10038 for request flush **
BlockList 93: in Bypass, line 20
<start> 171: Java @ Start
<start> 171: Kill window manipulation
<start> 171: Kill java functions
<end> 171: Kill event handlers
<end> 171: Java @ End
Edited by - MorpheusDreamlord on 13 Jun 2002 18:31:05
Edited by - MorpheusDreamlord on 13 Jun 2002 18:50:17
Post by: altosax on June 13, 2002, 06:07:18 PM
i've tryed the link provided above and no pup-up appeared but when i tryed to access to the domain i experienced a system block and had to reboot. this happened 3-4 times in the last 2 days with naoko 4.3, win me, ie 5.5 in different web page in a random way. i'm sure that i haven't software conflict because i have recently formatted my system and still not have the time to install the rest of the software.
i'm not using the new proxo commands, byte limit is under 16000 in each filter and never had problem with naoko 4.2, that i used until yesterday with the same filter set. note also that i have this forum in the bypass list and yesterday it happened a system block also here!
if i will be able to reproduce the problem i'll post here the step to follow.
regards,
altosax.
Post by: altosax on June 13, 2002, 06:10:04 PM
altosax.
Post by: MorpheusDreamlord on June 13, 2002, 06:15:47 PM
Please check Proxo43, please... I just had a major ifshlp.sys bsod, just after restarting prox43 and going to webnacks again
(that popup page is a great test).
http://www.webknacks.com/aptest_c1.htm
This has never happened under 42, and all other settings (explorer, IE, etc) are the same as last month, and so on back.
I truly have had a lot more system errors after using 43 than 42 (to be honest I have had some errors before 43, but not this many, and so severe - usually they are of the 'unstable, but recoverable' error... now they are 'no bsod, just reboot' type.
Post by: altosax on June 13, 2002, 06:50:30 PM
altosax.
Post by: JD5000 on June 13, 2002, 08:28:07 PM
quote:
Unfortunately many people are using configs that do all kinds of unusual things, and all I can vouch for is the default set.
My config doesn't do anything too unusual.. *evil grin*
LoL--------
"Imagination is more important than knowledge" - Einstein
Post by: Scott Lemmon on June 13, 2002, 11:06:47 PM
Since the page is a JavaScript exploit test, I'd suspect it may be IE causing the BSOD - its has some pretty low-level hooks (all that OS integration). The new version of Proxomitron uses more HTTP/1.1 features which may expose IE or TCP/IP bugs that 4.2 did not.
The usual suspect for driver problems are firewalls, virus scanners, and video drivers, but even those built into windows can have flaws. WinME can be a software conflict all by itself.
Make sure you have all patches for ME and IE installed first.Post by: Scott Lemmon on June 13, 2002, 11:45:08 PM
For a harsher test I went to slashdot.org and loaded up some of the larger articles. First thing I noticed it doesn't seem to like the super opener much - really messes up the layout, but the page loaded every time (and I reloaded a bunch). This is on Win98, so not even as crash proof as Win2k or XP is supposed to be. Never once did I get a crash, let alone a BSOD.
I'm not try to say it's not happening to people, but there must be some common factor for this problem that just isn't part of my system. All I can suggest is try different things like...
Does it happen in a different browser?
(try opera or mozilla for example)
Does it happen in the default.cfg too?
Does it happen if you disable persistent local connections?
(look in the config dialog under HTTP)
Does it happen if you make the timeout longer/shorter?
Is the log window open or closed?
Does it happen if web/header filters are bypassed?
Are you using an external proxy?
Also what browser & version are you using?
What OS?
What (if any) other programs are running?
Remember if you want to see this fixed, then the more useful information you can provide the better.
Regards,
Scott
Post by: ridgydidge on June 14, 2002, 12:56:30 AM
The page reload problem does happen in the default set that comes with 4.3. It happens less often but it did just happen loading this site. It happens spasmodically, mainly, but not exclusively, in forum sites like this. I use xp pro, IE6 (skinned with MyIE 3.3) with Outpost as my firewall and Allegrosurf as cache controller. Change back to the mighty Proxo 4.2 the problem goes away.
Yay, back to using JD5000 for cruisy surfing, but give me a little time and I'll reload 4.3 and try out some of the other options
Post by: pooms on June 14, 2002, 02:06:19 AM
at work and at home and have had no serious problems. I've tried every
site that anyone on this board or the yahoo groups has said gave them
problems, and I've seen no problems at all. I described my computer
in another post (W2K Server & IE6). In addition I'm using ZoneAlarm,
although I'm also behind a NetGear router with firewall at home, and
a linux iptables-based firewall at work. And I'm more or less using Jor's
filter set.
Post by: Scott Lemmon on June 14, 2002, 02:15:59 AM
Another question I just thought of - if your filters tack something on the end (like the toolbar in the JD5000 set) does it appear on the half-loaded page? That might tell me if it's getting cut off between the webserver and prox or prox and the browser. Although even that's not 100% since the browser may not render the ending HTML if some of the tags and aren't closed properly. It's possible if some middle section of the page went missing, the rest of the page might appear blank (esp. if it broke in the middle of a tag).
Also IE & Netscape have problems where, sometimes when you view source, you'll always see the first page load - even if following reloads are different (one of my big motivations for adding the view source option).
Post by: Scott Lemmon on June 14, 2002, 02:21:02 AM
quote:
I described my computer in another post (W2K Server & IE6).
Hm, maybe scratch IE6 off the list then.
I've since been surfing in IE5 with the JD5000 set. Gone to about 50 different sites - especially ones with big message boards, lots of HTML and images, etc. So far no problem.
For the descriptions it seems the cut-off usually happens while waiting for more data to arrive. Could mean something somewhere is timing the connection out.
Post by: Jor on June 14, 2002, 02:41:15 AM
quote:I don't use MSIE, but it happened in Opera 6.03, 6.04, Mozilla 1.0, and Moz 1.1a. I never use your default.cfg, only my own set (and test filters).
Does it happen in a different browser?
Persistent local connections/timeout longer/shorter: never touch those settings... do have HTTP/1.1 and pipelining enabled.
I usually have the log window close... never had a crash with it open.
Never happened in bypass mode yet, and not using any external proxy.
Browsers: Opera 6.04 and Mozilla 1.1alpha.
WindowsXP professional, and the crashes only occur with Kerio loaded.[/quote]
Post by: Jor on June 14, 2002, 02:43:32 AM
quote:Yes, always. No error messages from Proxomitron or the browser even, but the source is definately only partially there.
Another question I just thought of - if your filters tack something on the end (like the toolbar in the JD5000 set) does it appear on the half-loaded page?
Post by: Jor on June 14, 2002, 03:08:05 AM
(Web):Tag manager, Alt attribute to title, and Style links
(Headers):Cache-Control removed: private, If-Modified-Since removed, and If-None-Match removed.
The page stopped loading IMMEDIATELY after a request was made for Local.ptron/prox.css (a CSS file I always include), log excerpt:
Match 1960: Alt attribute to titleCan it be a (succesful) request for bweb..local.ptron/prox.css also causes the "parent" document to close sometumes?
BlockList 1961: in Bypass, line 28
+++GET 1961+++
GET /prox.css HTTP/1.1
User-Agent: Opera/6.04 (Windows; U)
Host: local.ptron
Accept: text/html, image/png, image/jpeg, image/gif, image/x-xbitmap, */*
Accept-Language: en
Accept-Charset: iso-8859-1;q=1.0, utf-8;q=1.0, utf-16;q=1.0, iso-8859-1;q=0.6, *;q=0.1
Accept-Encoding: deflate, gzip, x-gzip, identity, *;q=0
Referer: http://asp.flaaten.dk/pforum/
If-Modified-Since: Mon, 10 Jun 2002 14:19:28 GMT
Cache-Control: no-cache
TE: deflate, gzip, chunked, identity, trailers
Connection: keep-alive
+++RESP 1961+++
HTTP/1.1 304 Not Modified
Date: Fri, 14 Jun 2002 01:59:49 GMT
Server: Proxomitron
Connection: close
Content-Length: 0
<end> 1960: Mark <end> 1
<end> 1960: Print URL in page
<end> 1960: Tame Javascript (part 2)
+++CLOSE 1961+++
<end> 1960: Mark <end> 2
+++CLOSE 1960+++
Post by: Scott Lemmon on June 14, 2002, 03:29:18 AM
quote:
Can it be a (succesful) request for bweb..local.ptron/prox.css also causes the "parent" document to close sometumes?
Only if the browser decides to close it. Each connection happens in a different thread, and they're independent of each other as far as Proxomitron's concerned.
Does it happen if you don't load the file? Also try playing with the persistent connection settings - that's really the only major change since 4.2 that might affect a browser.
Post by: MorpheusDreamlord on June 14, 2002, 04:40:10 PM
Not disputing you (anyone good enuf to make Proxo in the first place knows his oats!!
But (there's always a butt)...
I've also personally seen (and read/heard of) a number of programs that, for various reasons (bad mem address/bad writes/etc) will trash a hard drive/currput memory/reboot windows (untrapped overwrite memory) (That game from 3 yrs ago is one, forgot it's name).
Proxo seems wonderful if you leave it alone, and just browse.
If you tick and untick filters, save, load, reload, etc the cfg file - just generally faff about, after a while (30 mins or so) I guarantee a crash.
I can repeat this anytime, any website - just keep faffing about with Proxo window opened (main, not log).
|
Come to the Dreaming...
Post by: Scott Lemmon on June 14, 2002, 11:58:04 PM
It's quite random and may only crash a program unlucky enough to be in the right spot in memory. In fact the driver may not crash at all - only the other program. Normal user-level programs (where Proxomitron runs), are not allowed to write to memory they don't own. Doing so should only cause the program to crash but nothing else.
If a driver has a bug, a program may make a perfectly legal call that the driver intercepts, but if it handles it improperly you may get a BSOD or other types of crashes. The crashes may happen in the driver, in the program, or even some completely unrealted program running at the time. There's often nothing the calling program can do - errors like this can go unnoticed until some software is unlucky enough to make a call in just the right way or at the right time.
When you get a BSOD the module it reports is where the crash happen, that's often (but not always) the driver at fault. Still, any driver can corrupt other drivers too. Other possible causes are usually hardware related (bad memory, overclocking too much, flakey mother board, etc.). Even these hardware errors may seem to only affect one program - at least for awhile, but just like with drivers, it's a crash just waiting to happen.
This is all I'm going to say on the matter. Please either accept what I'm saying or just don't use Proxomitron. There's probably nothing I can do to help your crashes, and I can't offer "official" support for any PC other than my own.
Post by: Arne on June 15, 2002, 12:08:15 AM
Best wishes
Arne
Imici username= Arne
Post by: Jor on June 15, 2002, 01:38:38 AM
I've switched to another firewall for now, no more crashing problems.
As for the partially loading pages -- I can live with that... reloading fixes it. Maybe I'll find what is causing it sooner or later, for now back to normal browsing with 4.3!
Post by: Scott Lemmon on June 15, 2002, 03:26:44 AM
Thanks!
Post by: JD5000 on June 15, 2002, 07:31:46 AM
http://home.satx.rr.com/jd5000/images/half.jpg
I went through my config filter by filter.. & I was able to narrow it down to "Insert Style Sheet - Links {1.d.se}", which inserts my css file into every page.
When I disabled this filter, I was able to load the page 30 times without the half loaded page appearing. Right after this, I enabled it, which caused the half loaded page to appear 7 out of 10 times. Just to make sure, I disabled it again & I was again able to load the page numerous times without error.
I changed the filter to add the styles directly, instead of calling a local file containing the styles. So far, on 20 page refreshes the half loaded problem has not returned.
I dunno if this will help... but.. just incase.
~JD
--------
"Imagination is more important than knowledge" - Einstein
Edited by - JD5000 on 15 Jun 2002 08:34:49
Post by: JD5000 on June 15, 2002, 07:45:44 AM
W98
Opera6
Kerio
--------
"Imagination is more important than knowledge" - Einstein
Post by: Scott Lemmon on June 15, 2002, 09:00:27 AM
Also, didn't someone report it still happening even with the default filter set? It does seem Kerio may cause some problems, but according to Jor it may not be the culprit here either.
BTW I still haven't heard from anyone about different settings for the local persistent connections and timeouts yet. I'd *really* like to know if these have any effect on the problem.
Edited by - Scott Lemmon on 15 Jun 2002 10:01:40
Post by: Arne on June 15, 2002, 09:05:44 AM
Best wishes
Arne
Imici username= Arne
Post by: Scott Lemmon on June 15, 2002, 09:29:17 AM
This would have no effect in Proxomitron 4.2 since without local persistent connections, you can't pipeline. It still wouldn't explain problems in Opera thou. It pipelines by default I believe, and I haven't had any similar troubles when testing it.
Also I understand that you don't normally use the default set and haven't altered the timeouts, but I do need these things tested too.
Edited by - Scott Lemmon on 15 Jun 2002 10:31:53
Post by: JD5000 on June 15, 2002, 10:04:31 AM
When you say "alter timeouts", do you mean the "local connections persist for xx seconds"? If so, I did try setting it to 30 seconds (with no luck) before I started messing with the filters. Does http 1.1 have to be enabled to pipeline? If so, unless they changed it again, Opera by default is set to http 1.0 for proxies.
--------
"Imagination is more important than knowledge" - Einstein
Edited by - JD5000 on 15 Jun 2002 11:06:02
Post by: cj. on June 15, 2002, 03:52:46 PM
I too have received a BSOD on my W2K Pro which is a very stable install, and as I had rembered the Kerio usage as the cause I couldn't help but notice that this particular "Stop error" message referred to the "fwdsys.drv": which of course is Kerio.
I was in the midst of a redir and http://www.newsnow.co.uk when it had occured.
I shall keep an eye out again and come back an note any further details. Please note that I have had Kerio ver 2.1.2 running without a hitch since it's install, and yes I am aware of the upgrade now available.
Post by: Jor on June 15, 2002, 05:01:25 PM
spent half an hour browsing with the default.cfg on various sites I know to cause problems, nothing seems to be wrong. That seems to narrow it down to something in my config, right?
Next I tried disabling 'Local connections persist' in Proxomitron's GUI, this didn't help. Got the partial page errors immediately.
Setting them to 30, caused Prox to generate error messages, and die.
Currently browsing with 15 -- no half-loaded pages yet (been browsing 30 mi nutes
).Also disabled Mozilla's pipeling (but haven't used that browser yet, it is my backup for DHTML sites). Do you know if Mozilla 1.1alpha has better pipelining?
Also, I have HTTP/1.1 enabled for proxies in Opera. Default is disabled.
Edited by - Jor on 15 Jun 2002 18:31:49
Post by: Scott Lemmon on June 15, 2002, 05:20:43 PM
There might be some difference in a way a filter is working between the two versions of Proxomitron - if you can narrow it down at all that would help.
Also are you saying seting the connection timeout to 30 caused crashes? That shouldn't be - I normally run it at 60 or higher.
Edited by - Scott Lemmon on 15 Jun 2002 18:22:57
Post by: Jor on June 15, 2002, 05:28:58 PM
[Edit] But setting it to 30 or over brings the partially loading pages back
[/Edit]As for Opera's pipelining -- it is enabled by default, when HTTP/1.1 is used (always when not using a proxy, since 6.01 disabled for proxies by default).
To enable pipelining (and other HTTP/1.1 behaviour), Opera6.ini has to be edited, and this string added:
[Proxy](Not yet documented at http://www.opera.com/support/mastering/operaini/ ).
Enable HTTP 1.1 for proxy=1
Also, as JD5000 said, the only filter which appears to be involved with this problem, is the 'Insert Prox stylesheet' filter (which in my case includes
<!--//--><link href="http://bweb..local.ptron/prox.css"on <start>.
type="text/css" rel="stylesheet" title="Prox CSS" />
$STOP()
Edited by - Jor on 15 Jun 2002 18:32:41
Edited by - Jor on 15 Jun 2002 18:43:04
Post by: Scott Lemmon on June 15, 2002, 05:39:36 PM
quote:
Also disabled Mozilla's pipeling (but haven't used that browser yet, it is my backup for DHTML sites). Do you know if Mozilla 1.1alpha has better pipelining?
I haven't downloaded it yet, but I've been a bit worried lately - recent versions of Mozilla seems to have *more* networking bugs than before. Several have been reported but don't seem to be getting fixed. For example, I notice at Google and Paul's ComputerCops forum Mozilla will often completely load the page twice before rendering it (very annoying because it also happens when posting sometimes). I first noticed when doing tests with the ALERT command - I had it prompt on each page load.
If you watch Mozilla's progress bar you can see it expand twice too, and it also happens when no proxy is in use. Only at specific sites though - it's as if some bit of HTML on the page triggers it. I was reading some messages about this to the developers who seemed not to believe it was happening, but it's hard to notice unless you have something watching the connections.
Post by: Jor on June 15, 2002, 05:45:36 PM
Post by: Scott Lemmon on June 15, 2002, 06:58:54 PM
If you increase the waiting period, that just means the connection is more likely to be reused with future pages, but it depends how long you wait between loading them. What happens if you turn on "Misc. info" in the log window? It'll show when connections are created and die. Try setting the time-out to 1000 or so to maximize reuse and try loading a bunch of pages. You should see connections being reused - including CSS requests (BTW I figure the browser shouldn't even ask for the CSS on every page - it should be cached). Also if you have your filters only add your CSS sheet does it still happen (just to rule out other filter interaction)? Thanks!
Post by: Jor on June 15, 2002, 09:06:53 PM
I tested the following:
quote:
if you have your filters only add your CSS sheet does it still happen (just to rule out other filter interaction)
It does. I am fairly sure I found what causes the problem: I removed $STOP() from my include Prox CSS filter (see post 2297 in this thread), and no more partially loaded pages. I have no idea why this makes all the difference, but I can load all pages completely now with any value for 'Local connections persist' in Proxomitron, even disabling it!
quote:
(BTW I figure the browser shouldn't even ask for the CSS on every page - it should be cached).
Opera6 completely follows the W3C recommendations on HTTP Headers, so reloading the current page (with F5 or the mouse gesture right-click-hold up-down) also requests all images and other linked files on that page. Since Prox serves files with the HTTP Header 'Cache-Control: no-cache', these will always be completely reloaded, effectively bypassing the cache.
Edited by - Jor on 15 Jun 2002 22:08:58
Post by: TEggHead on June 15, 2002, 09:23:18 PM
quote:
I usually keep the log window open (I love the new logging options, now if you could find a way to automatically write it to a text file?)
eheh
$ADDLST(LOGfile,[$DTM(d T)] | | u | )HTH
JarC
Edited by - TEggHead on 15 Jun 2002 22:24:58
Post by: MorpheusDreamlord on June 16, 2002, 05:25:44 AM
??
same with <end> ??
|
Come to the Dreaming...
Post by: altosax on June 16, 2002, 11:22:45 AM
i've asked this exact question to scott at computer cops and both tegghead and scott have answered this.
after their answers i've wrote it in my tutorial, now updated in the p-faq section.
regards,
altosax.
Post by: JD5000 on June 16, 2002, 11:28:37 AM
--------
"Imagination is more important than knowledge" - Einstein
Post by: JD5000 on June 16, 2002, 12:32:48 PM
http://virgolamobile.50megs.com/proxomitron.html
And the half-loaded problem occurred.. So, I disabled the other filter that inserts a css & again the problem went away.
I tried setting the connection timeout between 1 - 60, but it had no effect. Just for fun, I changed the match from <start> to <end> and... it worked. LoL I doubt this helps any, but what the heck.
--------
"Imagination is more important than knowledge" - Einstein
Post by: altosax on June 16, 2002, 12:53:55 PM
why don't you insert the stylesheet in a more common way?
Match = "</head>|(<body)1$SET(2=<head>)"
Replace = "2"
"<your stylesheet here>"
"</head>"
"1"
"$STOP()"
regards,
altosax.
Post by: MorpheusDreamlord on June 16, 2002, 03:53:35 PM
|
Come to the Dreaming...
Post by: Arne on June 16, 2002, 04:12:02 PM
Best wishes
Arne
Imici username= Arne
Post by: MorpheusDreamlord on June 16, 2002, 04:38:39 PM
It happened to me on this page, ok. I reloaded - the page reloaded very fast (like from the cache), EVEN thought I did a shift-reload (IE force reload).
Still blank. Repeated shift-reload, repeat, repeat, still blank.
Empty cache and history (I always do both together..), reload. Page is fine, perfect even.
Is Proxo telling the browser to ignore the force-reload? Maybe the browser expects to find the page all in cache, and it's not all there....
(or am i talking outta my rear again...?)
|
Come to the Dreaming...
Edited by - MorpheusDreamlord on 16 Jun 2002 17:41:22
Post by: Scott Lemmon on June 16, 2002, 07:36:53 PM
If you're still using the WinME base install, you should probably do a windows update and get your system patched up.
Post by: Arne on June 17, 2002, 12:15:54 AM
Removing the "write to file" immediately fixed the problem.
Best wishes
Arne
Imici username= Arne
Post by: Arne on June 17, 2002, 12:21:22 AM
Added: I had to try this before I went to bed. And it is not related to 2 filters writing at the same time. Having only one filter that can write to the file is enough for this to randomly happen.
using IE6 & W98 all patches installed.
Best wishes
Arne
Imici username= Arne
Post by: sidki3003 on June 17, 2002, 02:29:51 AM
sidki
Post by: Scott Lemmon on June 17, 2002, 04:56:03 AM
Arne: ADDLST does locking so only one entry can write to the file at a time. Not sure how it could cause a half-loaded page though. What do the filters look like where you're using it?
Post by: Arne on June 17, 2002, 06:11:13 AM
Not a big deal, because I am not going to use it on filters that occur every time though. I was planning on using it in filters which I suspect very seldom kick in.
Best wishes
Arne
Imici username= Arne
Post by: MorpheusDreamlord on June 17, 2002, 08:15:42 AM
quote:
Sounds like you're version of IE still has the Last-Modified: cache bug. I removed the work-around Proxomitron had for that since it was causing other problems.
If you're still using the WinME base install, you should probably do a windows update and get your system patched up.
Yup, got the latest May update (unless there's a later one..)
|
Come to the Dreaming...
Post by: Multiproxomatrix on July 11, 2002, 10:22:20 AM
I watched this thread dealing with the partial page loads when it was active very closely, but I didn't want to chime in until I knew anything definite. I was getting the partial loads with such frequency that I returned to using 4.2, and began examining the programs differences. The only apparent thing I saw that would cause {Scott, you would know for sure :)}this was the new connection persistance feature. I returned to 4.3, but unticked "local connections persist". I haven't gotten a partial page load in two weeks with constant use. I think there must be a problem in the way the local persistance is being implemented. Hope this helps.
Multiproxo
IE6 build 2600
Windows 2000 SP2
Sygate 5 build 1120
Mainly using JD's config at this time with a few small additions.
Post by: dave1006 on July 11, 2002, 12:59:20 PM
I only ever got the half page loads when using pipelining with Mozilla... (Which was|is buggy).
With Opera and IE, and Mozilla with proxy pipelining off, I never have any problems at all with Proxo.... It runs as beautifully as ever
dave
dave at smokeajay.co.uk
Post by: Scott Lemmon on July 12, 2002, 04:27:29 PM
They've also fixed another annoying bug that would randomly cause some pages to load twice - it could be hard to notice, but on some pages you'd see two requests in Proxomitron's log window. It could cause double posts at some sites too.
As far as the persistence goes it works fine, but you may be having the local file problem described in the other posts. Disabling local persistence will avoid that, but using the filter should work better.
Funny that there's not much consistency in the symptoms people describe as causing this - for example, some were reporting turning off local persistence actually made the half-loads worse. Since many things - both Proxomitron related and not - can cause partial page loads, I suspect it may not all be the same issue.
Post by: Multiproxomatrix on July 16, 2002, 08:51:49 AM
, so I don't miss not using it in 4.3. The program is fantastic. Good luck with fixing it. I'm not sure what you do for a living, but thank you for your efforts, and for finding the time to give us this vital piece of software.M.P.M.
Post by: xartica on August 02, 2002, 01:54:52 AM
THanks! Unticking "local connections persist" solved the half-laoding page problem for me too.