aMule Forum
English => aMule Help => Topic started by: esteryl on February 08, 2005, 08:50:45 PM
-
I recently updated amule rc7 to rc8 and now, when i make a search using amuleweb. it displays results like:
File Name Size File Hash Sources Download
ýýýýýýýýýýýýýýýýýýýýý ýý ýýýýýýýýýýýýýà,3Something [spanish divx].mpg 27
ýýýýýýýýýýýýýýýýýýýýýýýýýý ýýýý‰ýýýý ýýýýýýýýýýýýýà,3Something [spanish divx].mpg 27
ýýýýýýýýýýýýýýýýýýýýýýýýýý ýýýýýýýý ýýýýýýýýýýýýýà,3Something [spanish divx].mpg
where "Something [spanish divx].mpg" is the name of something i'm already downloading.
Where's the problem?
I use debian and installed debian package from downloads page.
Thanks
-
Did you ever figure out what was wrong? I still see this. I built from the source in case it was a problem with the debian packages but this didn't help.
Amuleweb broken!! :(
-
did you guys update the aMule.tmpl in your .aMule dir?
amuleweb is NOT brocken in rc8 ;)
-
Hello !
I had the same problem.
I replaced my old aMule.tmpl by the new in the cvs, and I din't have this strange words.
Now it's blank... ;(
Now when I'm performing a search, I have this result:
[FILENAME] [FILESIZE] [SOURCECOUNT]
[FILENAME] [FILESIZE] [SOURCECOUNT]
[FILENAME] [FILESIZE] [SOURCECOUNT]
[FILENAME] [FILESIZE] [SOURCECOUNT]
[FILENAME] [FILESIZE] [SOURCECOUNT]
but, no name in filename, no size in filesize...
I use the rpm'version for Mandrake 10.1...
Please help me....!
-
totoetlititi, if you use rc8 you have to take the aMule.tmpl from the rc8 sources. amule, amuleweb and aMule.tmpl must always be the same version!
-
I replaced these files with the file in my rpm, and the pb is come back...:
The result of a search is:
ýýýýýýýýýýýt8!ý ýýa Not declared ýýýýýýýý 689
ýýýýýýýýýýýýý> ýýa Not declared ýýýýýýýý 541
ýýýýýýýýýýýý8Iýýýý ýýa Not declared ýýýýýýýý 380
ýýýýýýýýýýýt8!ý ýýa Not declared ýýýýýýýý 315
Why ? ;(
-
totoetlititi,
This looks like UTF-8 chars displaying. You can try this:
$ LANG=en_US.UTF-8 amule &
$ LANG=en_US.UTF-8 amuleweb -f
Tell me if this makes a difference.
-
Thanks, but no difference...
But, my amule is always in french after LANG=en_US.UTF-8 amule & ...
Is it a problem ?
[EDIT]
I'm running amule 2.0.0rc8 using wxGTK2 v2.5.3 (Unicoded)
[EDIT]
-
Well, it is quite strange, should be in english. Type:
$ env | grep LC
$ env | grep LANG
Maybe another locale setting is prevailing. Btw, if you want to test, you should not be using rc8, there were so many bug fixes in amule/amuleweb, that I can't count. Use a CVS snapshot, get one here:
http://www.hirnriss.net/?area=cvs
And read this:
http://www.amule.org/wiki/index.php/Backtraces
Cheers!
-
Thanks !
I finally find the ultimate solution...
(no, it's not emule.... :])
I re-installed amule, but not the unicode version, only the ansi version, and that's good !!!
:baby: amule...
-
Thank you for taking the time to help!
I had already matched the aMule.tmpl and webserver directories to the sources I built from.
I also investigated heavily whether locale was somehow affecting the search feature, trying all kinds of different things, starting with what you suggest (it is also my default locale). Nothing made any difference.
Then I rebuilt from the latest CVS. This was challenging. I finally figured that there is a dependency on wxWidgets CVS features later than 2.5.3, so I built that as well to solve the missing archive.h issue.
That got search working, except that the CVS tree is unstable for me.
Very frequently amuled seems to be wedged on startup and need to be restarted, with no output in the logs...
Well, it is always talking about trying to connect to "0.0.0.0 : 0" - which also shows up in the server list no matter how often it's deleted. Maybe that's a clue.
It could be coincidence, but I had a sense that if I delete server.met it does start, and run for a while (hours?) before it gets wedged like that again.
Well, it's a CVS version, so whatever. :]
-
Originally posted by tsark
Then I rebuilt from the latest CVS. This was challenging. I finally figured that there is a dependency on wxWidgets CVS features later than 2.5.3, so I built that as well to solve the missing archive.h issue.
That's not really correct, the file archive.h is there in 2.5.3, only somebody at debian screwed up while building the package and this file is missing in the "debianized" wx... There's a sticky in the compilation problems forum with an adress where to download the missing file if you want to go back to debian wx 2.5.3
-
Ahh!
That's funny. I googled it and got a reference from some site with wxWidgets "v2.5.4" APIs, so I just assumed that was it.
Thanks for that. Yeah, no reason to have two copies of wxWidgets sitting around.
FYI amuled is on a roll, aMule-cvs-20050220.tar.bz2 has now been running for 24 hours+...
-
Originally posted by tsark
FYI amuled is on a roll, aMule-cvs-20050220.tar.bz2 has now been running for 24 hours+...
Great! Keep testing and reporting, so that next release will be the most stable ever! ;)
Cheers!
-
Too bad, it is still wedging.
;(
This time it went almost 30 hours, but then I ended up with the same thing, "Now connecting" and everything really slow or not responding at all.
-
tsark,
Please state your problem clearly now, I can't understand it from the previous messages. Tell me exactly what is happening.
-
I've been lookng kind of carefully at this for a few days. Here's a summary of what I've seen so far. Note that I start with a blank slate, no existing .aMule dir.
*) Periodically (no pattern I can see yet) there is nothing visible in the log or debug log (via the web). "WARNING: logfile is empty. Something wrong" Sometimes when you start it up it works, other times not... More often not. When it does work the last line often truncates in the middle, often with a garbage character. Buffer problem? (100% repeatable)
*) Settings in preferences (number of connections) appear to be ignored, or at least not saved. I adjust things by hand via .eMule. (100% repeatable)
*) server.met downloaded via the web interface seems not to work. When amuled starts up and I connect via amuleweb, I see a correct-looking list of servers, but status is "now connecting" and it never connects. stdout indicates repeated attempts to connect to "0.0.0.0 : 0". Very quickly if I click around, the web interface will stop responding. Once it's wedged, amuleweb starts spitting out socket errors. (recreated this twice)
*) If I rm the server.met file, everything is stable but with no servers. If I hand-enter a single server, everything works more or less OK for a while, which is what I've been doing. But very quickly a server appears in the server list: "0.0.0.0 : 0"; you can remove it but it always comes right back. (100% repeatable)
*) Eventually after 10-30 hours I'll check the web interface and see that dreaded "Now connecting" message. It never goes away. The web interface will be slow at that point and will eventually stop responding. (This has happened twice)
*) Weirdest thing of all. After running for some number of hours, I run top on the box, and see a load average of 3.0-5.0 and 75% user CPU, 25% sys. Meanwhile no processes actually in the list are showing up as taking any CPU time! By the summary the box looks swamped, but by the process list it looks idle! Crazy. Occasionally amuled blinks up to the top and the most it ever appears to use is 5% CPU. Meanwhile the system is brutally sluggish and obviously struggling.
That's crazy, watching linux bog down while not indicating what process is doing it, seeing the two stats disagree with each other. Never seen that before. Sure enough, on a hunch I kill amuled and instantly everything is normal again. (happened twice)
-
Originally posted by tsark
*) server.met downloaded via the web interface seems not to work. When amuled starts up and I connect via amuleweb, I see a correct-looking list of servers, but status is "now connecting" and it never connects. stdout indicates repeated attempts to connect to "0.0.0.0 : 0". Very quickly if I click around, the web interface will stop responding. Once it's wedged, amuleweb starts spitting out socket errors. (recreated this twice)
*) If I rm the server.met file, everything is stable but with no servers. If I hand-enter a single server, everything works more or less OK for a while, which is what I've been doing. But very quickly a server appears in the server list: "0.0.0.0 : 0"; you can remove it but it always comes right back. (100% repeatable)
*) Eventually after 10-30 hours I'll check the web interface and see that dreaded "Now connecting" message. It never goes away. The web interface will be slow at that point and will eventually stop responding. (This has happened twice)
This is solved on current CVS. New tarball will follow soon.
Originally posted by tsark
*) Periodically (no pattern I can see yet) there is nothing visible in the log or debug log (via the web). "WARNING: logfile is empty. Something wrong" Sometimes when you start it up it works, other times not... More often not. When it does work the last line often truncates in the middle, often with a garbage character. Buffer problem? (100% repeatable)
This seems to come from unicode issues. Will look at it later.
Originally posted by tsark
*) Settings in preferences (number of connections) appear to be ignored, or at least not saved. I adjust things by hand via .eMule. (100% repeatable)
Bad thing. I'll check it also.
Originally posted by tsark
*) Weirdest thing of all. After running for some number of hours, I run top on the box, and see a load average of 3.0-5.0 and 75% user CPU, 25% sys. Meanwhile no processes actually in the list are showing up as taking any CPU time! By the summary the box looks swamped, but by the process list it looks idle! Crazy. Occasionally amuled blinks up to the top and the most it ever appears to use is 5% CPU. Meanwhile the system is brutally sluggish and obviously struggling.
That's crazy, watching linux bog down while not indicating what process is doing it, seeing the two stats disagree with each other. Never seen that before. Sure enough, on a hunch I kill amuled and instantly everything is normal again. (happened twice)
Probably a threads issue. tried ps to see amuled threads?
-
Wow, you guys are fast! :baby:
I just now read about ps's ability to give thread info, after you mentioned it. Next time it happens I will use it and see what it looks like.
Strange, I would have thought a process' threads' CPU time would be counted towards the process, but who knows what's going on there really.
I will do another build from CVS soon, at latest over the weekend...
-
Latest progress:
*) Server connection issues appear resolved. After 4+ days of testing, I haven't seen the dreaded "0.0.0.0 : 0" reappear once.
*) I am still seeing amuled pegging the CPU at 100%, load averages of 3-5. This looks the same, 75% user 25% sys. On your advice I did ps auww -T and it is indeed one amuled thread doing it. It's unpredictable and difficult to reproduce. It feels like if I let it run for a while it will happen eventually. But once this morning I got it almost immediately, within seconds after starting the daemon.
*) Preferences changes still not saved.
*) Another minor detail, transfer and shared file page controls don't seem to work (i.e. adjusting priorities seems to have no effect, or - once - the opposite effect)
-
tsark,
What is the CVS version you are testing? The high cpu usage should have been fixed in the 20050221 tarball. Please, report if this behaviour persists and do post a backtrace if possible.
Thanks!
-
Originally posted by tsark
Latest progress:
*) Server connection issues appear resolved. After 4+ days of testing, I haven't seen the dreaded "0.0.0.0 : 0" reappear once.
Good to know
Originally posted by tsark
*) I am still seeing amuled pegging the CPU at 100%, load averages of 3-5. This looks the same, 75% user 25% sys. On your advice I did ps auww -T and it is indeed one amuled thread doing it. It's unpredictable and difficult to reproduce. It feels like if I let it run for a while it will happen eventually. But once this morning I got it almost immediately, within seconds after starting the daemon.
Will be nice to know what'a happening there, indeed.
Originally posted by tsark
*) Preferences changes still not saved.
Fixed now.
Originally posted by tsark
*) Another minor detail, transfer and shared file page controls don't seem to work (i.e. adjusting priorities seems to have no effect, or - once - the opposite effect)
Hum, GonoszTopi, lfroen, please?
-
Originally posted by tsark
*) Another minor detail, transfer and shared file page controls don't seem to work (i.e. adjusting priorities seems to have no effect, or - once - the opposite effect)
Hum, GonoszTopi, lfroen, please?
Will have a look.
-
adjusting priorities seems to have no effect, or - once - the opposite effect
I can confirm it did strange things.
Now it is fixed, will be in tomorrow's CVS tarball.
-
Originally posted by phoenix
tsark,
What is the CVS version you are testing? The high cpu usage should have been fixed in the 20050221 tarball. Please, report if this behaviour persists and do post a backtrace if possible.
Thanks!
After watching this for several days, I can confirm that amuled still ends up spinning and using up all CPU time eventually. This happens every time it is started. It may take minutes or hours but it appears 100% reliable.
I have been using aMule-cvs-20050225.tar.bz2.
Since this is not a crash, I don't know how I can get you a backtrace.
I suppose you'll want some way of knowing what code is spinning. If there is some way I can turn on more debug logging we can hope that there is a log message inside whatever code is hot, and then it'll show up in the logs. Or if there was a c/c++ code profiling tool we could use that to see where in the code all the time is being spent. Or perhaps there is some way to use GDB... if we could identify the busy thread, break its execution, and look at a stack trace?
Or perhaps the 75% user/25% system thing is a clue. Whatever's running over and over is making a lot of system calls. I could try running amuled in strace but man, I don't know if my disk is big enough to fit everything until the problem starts.
-
tsark,
Yes, you can run it inside gdb. It should be easy, just read these:
http://www.amule.org/wiki/index.php/Using_gdb_and_valgrind
http://www.amule.org/wiki/index.php/Backtraces
-
Originally posted by phoenix
tsark,
Yes, you can run it inside gdb. It should be easy, just read these:
http://www.amule.org/wiki/index.php/Using_gdb_and_valgrind
http://www.amule.org/wiki/index.php/Backtraces
Running it inside gdb I can do. Figuring out how to tell you anything useful from that I'm not so sure about.
I'm doing all Java development lately and my unix C is rusty. I have used gdb but not valgrind. Can valgrind somehow magically break execution in the problem area?
-
no, but you can break amued when you see that happen on gdb, list threads, and change to the needed one, then step inside to see what's repeating.