aMule Forum

Please login or register.

Login with username, password and session length
Advanced search  

News:

We're back! (IN POG FORM)

Pages: 1 2 3 [4]

Author Topic: strange ID behaviour  (Read 23381 times)

badjob

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 6
Re: strange ID behaviour
« Reply #45 on: September 15, 2005, 05:49:57 PM »
lionel77, could you please post or tell about the patches you did to wxWidgets? I certainly would like to try these on my CVS version of aMule. It seems the stuck connections are gone on my Tiger system as of CVS_2005-09-08.
Logged

lionel77

  • Provider of Mac builds, Forum Mod
  • Hero Member
  • *****
  • Karma: 4
  • Offline Offline
  • Posts: 1107
  • Mac OS X 10.4 (Power Mac G5)
Re: strange ID behaviour
« Reply #46 on: September 16, 2005, 06:29:40 AM »
badjob, did you check the number of stuck connections using netstat as described above? maybe you are just not reaching the limit for the problem to surface.

amule cvs has not been working with 10.3 for a while now (a number of weeks?) -- it crashes on average every 3-10 hours. i've posted a couple of backtraces but the problem has not been fixed yet.
originally, i was planing to release a binary that would work on both 10.3 and 10.4 but i'm starting to think i should just go ahead and post a patched 10.4-only binary.
if everything goes well you can expect that binary within 24h.
Logged
Current aMule CVS builds for OS X can be found here.

badjob

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 6
Re: strange ID behaviour
« Reply #47 on: September 16, 2005, 10:45:41 AM »
lionel77,
some time ago aMule always stopped working after some hours as some other people wrote in this forum. I got a lowid after some time and had stuck connections (Mac OS X 10.4.2).

Right now I'm running wx2.6.1 and aMule CVS head with major custom GUI patches (mine). No networking code was touched. I built the binaries with gcc4 with --disable-optimize --enable-debug and omitted the stripping.

No problems at all while having 50 up to 300 concurrent connections on average, depending on how much stuff I shared. This was for the standard monolithic GUI client.

Didn't use 10.3 on my main system since release of 10.4.

So a binary wouldn't help me that much, because of my custom GUI patches (some refinements), which I would gladly share in a couple of days when they are finished if someone is interested.
Logged

davenull

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 6
Re: strange ID behaviour
« Reply #48 on: September 21, 2005, 09:20:28 PM »
Hi,

Is there some news about the $@# problem concerning amule stopping using High ID after a given period of time ?

I'm using daily sources of amule cvs, build against wxMac-2.6.1 on Tiger 10.4.2 and i have to stop then restart amule each and every hour or so (when i have a High ID, ie on each start, the download rates are rather good).

I remember old versions of mlnet suffering of such a problem, but with a different effect: mlnet freezed and we had to kill the process to restart it again. I remember a script which periodically checked connexions and restart automatically mlnet if necessary. Could it be the same problem here ?
Logged

lionel77

  • Provider of Mac builds, Forum Mod
  • Hero Member
  • *****
  • Karma: 4
  • Offline Offline
  • Posts: 1107
  • Mac OS X 10.4 (Power Mac G5)
Re: strange ID behaviour
« Reply #49 on: September 22, 2005, 06:48:45 AM »
@badjob:
Again, just because you haven't reached the point where no more new connections are accepted does not mean you are not affected. Run amule for 2 days and then use netstat as described above. I would be surprised if you had no stuck connections by then.

The patch is rather simple. Open wxWidgets/src/unix/gsocket.gpp and go to line 533. It should look something like:
Code: [Select]
     (listen(m_fd, 5) != 0))change it to
Code: [Select]
     (listen(m_fd, 128) != 0))128 is the maximum allowed and it will result in amule tolerating up to 193 stuck connections.

Looking forward to your GUI patches... :)



@davenull:
The problem is that it is the underlying low-level os code that is causing the trouble. There is not much we can do about it, except for coming up with work-arounds like increasing the limit for stuck connections.
EVERY single p2p app I've tried so far is affected by this. They only differ in the time it takes until the problem surfaces. Mldonkey for instance tolerates 31 stuck connections per default while an unpatched amule tolerates only 8. Azureus (a BitTorrent client) tolerates 193, which is the maximum, and therefore needs pretty long until no more new connections are accepted.
Rumor is that 10.4.3 will contain a number of fixes on network code, so if we are lucky the problem will be history soon. If not, we have to flood Apple with bug reports... ;)

So in the meantime I suggest you try the most recent cvs binary I posted a couple days ago. It allows for 193 stuck connections instead of the default 8, so you don't have to restart amule that often.
Logged
Current aMule CVS builds for OS X can be found here.

davenull

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 6
Re: strange ID behaviour
« Reply #50 on: September 22, 2005, 11:37:07 AM »
Quote
Originally posted by lionel77
So in the meantime I suggest you try the most recent cvs binary I posted a couple days ago. It allows for 193 stuck connections instead of the default 8, so you don't have to restart amule that often.

Thanks for your answer.

Did you mean the last "SVN" binary ? (i didn't use it as you say in another post that this version is not that stable...).
Logged

lionel77

  • Provider of Mac builds, Forum Mod
  • Hero Member
  • *****
  • Karma: 4
  • Offline Offline
  • Posts: 1107
  • Mac OS X 10.4 (Power Mac G5)
Re: strange ID behaviour
« Reply #51 on: September 22, 2005, 04:39:14 PM »
SVN and CVS are both version control systems that help to keep track of the changes that are made to the code (from the user perspective it is completely irrelevant which one is used). Recently we switched from CVS to SVN so I started calling my binaries SVN binaries instead of CVS binaries. Then I realized that the rest of the team thought it was less confusing to stick with the CVS terminology, because many people don't know about SVN and already associate CVS with up-to-date code. So now I'm calling my builds CVS builds again, even though, technically, I retrieve the code through SVN.

More importantly, the aMule-Mac.cvs2005-09-12.zip binary I posted a few days ago should run ok under Tiger. The stability issues seem to be mostly limited to Panther.
Logged
Current aMule CVS builds for OS X can be found here.

davenull

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 6
Re: strange ID behaviour
« Reply #52 on: September 22, 2005, 04:51:44 PM »
Thanks for the explanations.

I've downloaded your cvs2005-09-12 binary and i'm testing now. Crossing fingers...
Logged

Mars_Artis

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 16
Re: strange ID behaviour
« Reply #53 on: September 24, 2005, 09:47:10 AM »
Ok ,one day of ful running of the lates CVS from you lionel, everything is working fine.
This is the result of netstat:

Mars-G5:~ mars$ netstat -aLp tcp | grep -w 17672
94/94/128      *.17672                
Mars-G5:~ mars$ netstat -ap tcp | grep "\*\.17672"
tcp4       0      0  *.17672                *.*                    LISTEN
Mars-G5:~ mars$ telnet localhost 17672
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
Logged

lionel77

  • Provider of Mac builds, Forum Mod
  • Hero Member
  • *****
  • Karma: 4
  • Offline Offline
  • Posts: 1107
  • Mac OS X 10.4 (Power Mac G5)
Re: strange ID behaviour
« Reply #54 on: October 01, 2005, 11:51:29 AM »
I've separated the postings on crashes and backtraces to keep this thread more legible. You can find the moved posts in  this thread. :)
Logged
Current aMule CVS builds for OS X can be found here.

samson

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 9
    • http://samson.net.ru
Re: strange ID behaviour
« Reply #55 on: October 01, 2005, 01:25:13 PM »
i need to comiple aMuledaemon (i've asked my question at this thread) for Mac OS X Server 10.4.2

so, do i need only to change this string: "(listen(m_fd, 5) != 0))" and than compile sources as usual? or i need to do something else?

sorry for my english. (-:
« Last Edit: October 01, 2005, 01:25:56 PM by samson »
Logged

lionel77

  • Provider of Mac builds, Forum Mod
  • Hero Member
  • *****
  • Karma: 4
  • Offline Offline
  • Posts: 1107
  • Mac OS X 10.4 (Power Mac G5)
Re: strange ID behaviour
« Reply #56 on: October 01, 2005, 07:33:35 PM »
Quote
Originally posted by samson
so, do i need only to change this string: "(listen(m_fd, 5) != 0))" and than compile sources as usual? or i need to do something else?
Yes, that's the only thing you have to change in wxWidgets for better Tiger compatibility.
Logged
Current aMule CVS builds for OS X can be found here.

samson

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 9
    • http://samson.net.ru
Re: strange ID behaviour
« Reply #57 on: November 09, 2005, 03:01:46 PM »
have anybody information about this bug in 10.4.3?
Logged

Kry

  • Ex-developer
  • Retired admin
  • Hero Member
  • *****
  • Karma: -665
  • Offline Offline
  • Posts: 5795
Re: strange ID behaviour
« Reply #58 on: November 09, 2005, 03:35:03 PM »
Is fixed on 10.4.3
Logged

samson

  • Approved Newbie
  • *
  • Karma: 0
  • Offline Offline
  • Posts: 9
    • http://samson.net.ru
Re: strange ID behaviour
« Reply #59 on: November 09, 2005, 03:46:13 PM »
in other words, i can compile it under 10.4.3 without any tricks with wxWidgets (like "(listen(m_fd, 5) != 0))") and it will correctly works?

great! (-:
Logged
Pages: 1 2 3 [4]
« previous next »