[mindpirate]

Major update of the Howto is now complete:

http://forum.emule-p...showtopic=42844

I'm a bit tired so there may be some mistakes. I'll look at it all again after a nice long sleep.

[hszFS]

Uhm, I just installed a fresh debian with the latest unstable packages and wine 0.9.5. eMule seems to work but if I look at $ top wine-preloader/wine-server is using 100% CPU all the time, even when eMule is started idle and not connected to a server.
Because of this, the webinterface also seems not to work (very sluggish).

Two questions:
1. Can anyone second that?
2. Does anyone have a clue what the reason of this might be?

Wine is set to Win98 correctly.

This post has been edited by hszFS: 08 January 2006 - 02:24 AM

[mindpirate]

hszFS, on Jan 7 2006, 09:21 PM, said:

Uhm, I just installed a fresh debian with the latest unstable packages and wine 0.9.5. eMule seems to work but if I look at $ top wine-preloader/wine-server is using 100% CPU all the time, even when eMule is started idle and not connected to a server.
Because of this, the webinterface also seems not to work (very sluggish).

Two questions:
1. Can anyone second that?
2. Does anyone have a clue what the reason of this might be?

Wine is set to Win98 correctly.

You should try reverting to 0.9.3 or 0.9.2, two recent versions of Wine that are known (at least to me) not to consume 100% CPU with eMule in Win98 mode as set up in the HowTo. Typical peak CPU usage with Kad + Server and a few downloads going is around 10-15% on my Athlon XP 3200+ box.

I'll upgrade to 0.9.5 myself and see if I can duplicate your problem. What version of eMule are you using BTW?

[hszFS]

mindpirate, on Jan 8 2006, 10:33 AM, said:

You should try reverting to 0.9.3 or 0.9.2, two recent versions of Wine that are known (at least to me) not to consume 100% CPU with eMule in Win98 mode as set up in the HowTo.

Uhm, I had 0.9.2 before and upgraded to 0.9.5 because of the CPU usage.

mindpirate, on Jan 8 2006, 10:33 AM, said:

Typical peak CPU usage with Kad + Server and a few downloads going is around 10-15% on my Athlon XP 3200+ box.

Well, the point is that it uses 100% even when I just it and it's completely empty and idle, not connected, nothing.

mindpirate, on Jan 8 2006, 10:33 AM, said:

What version of eMule are you using BTW?

0.46c of course.

[mindpirate]

I've upgraded to Wine 0.9.5 and I'm not able to reproduce your issue.

Short of running a profiler on eMule there's not too much you can do to discover which part of the code it's spinning in. If this is happening immediately after you launch eMule (and have run the first time setup wizard), my feeling is it might be when eMule is generating your Secure Identification key for the first time. That part of the code makes calls to the Windows CryptoAPI. Wine's implementation of CryptoAPI was just recently added and isn't well-tested yet--perhaps it's not finding a suitable source of entropy to generate your key.

To test this theory, with eMule shut down you should edit the file .wine/drive_c/Program Files/eMule/config/preferences.ini and disable functionality related to Secure Identification by changing the value of the following preferences to "0":

SecureIdent=0
LogSecureIdent=0

Save the file then launch eMule again and see if it still goes to 100% CPU.

[hszFS]

mindpirate, on Jan 11 2006, 03:03 AM, said:

I've upgraded to Wine 0.9.5 and I'm not able to reproduce your issue.

Short of running a profiler on eMule there's not too much you can do to discover which part of the code it's spinning in.

If there is a good and not too difficult to use profiler I will do that if you hint me out to one ;D

mindpirate, on Jan 11 2006, 03:03 AM, said:

If this is happening immediately after you launch eMule (and have run the first time setup wizard), my feeling is it might be when eMule is generating your Secure Identification key for the first time. That part of the code makes calls to the Windows CryptoAPI. Wine's implementation of CryptoAPI was just recently added and isn't well-tested yet--perhaps it's not finding a suitable source of entropy to generate your key.

To test this theory, with eMule shut down you should edit the file .wine/drive_c/Program Files/eMule/config/preferences.ini and disable functionality related to Secure Identification by changing the value of the following preferences to "0":

SecureIdent=0
LogSecureIdent=0

Save the file then launch eMule again and see if it still goes to 100% CPU.

Yes it is happening immediately when first starting a blank eMule client. I disabled SecureIdent but that did not help at all

[mindpirate]

hszFS, on Jan 22 2006, 07:53 AM, said:

If there is a good and not too difficult to use profiler I will do that if you hint me out to one ;D

If you're not an experienced programmer you shouldn't even bother messing around with a profiler.

Something else you could try is to upgrade to their latest versions all packages that Wine depends on.

[hszFS]

mindpirate, on Jan 23 2006, 11:19 AM, said:

hszFS, on Jan 22 2006, 07:53 AM, said:

If there is a good and not too difficult to use profiler I will do that if you hint me out to one ;D

If you're not an experienced programmer you shouldn't even bother messing around with a profiler.

I am a programmer. I just never messed with profilers. If its a matter of recompiling stuff and not modifying the code then I might try it.

mindpirate, on Jan 23 2006, 11:19 AM, said:

Something else you could try is to upgrade to their latest versions all packages that Wine depends on.

The Debian I used for testing is completely up to date. Does not help.

This post has been edited by hszFS: 24 January 2006 - 08:56 PM

[hszFS]

It also does not work with the latest beta version of eMule (0.47a beta2).

[mindpirate]

hszFS, on Jan 24 2006, 03:55 PM, said:

mindpirate, on Jan 23 2006, 11:19 AM, said:

Something else you could try is to upgrade to their latest versions all packages that Wine depends on.

The Debian I used for testing is completely up to date. Does not help.

Although you're using the Debian unstable repos, it's likely that at least one of your dependencies is a bit stale ("stale" in a relative rather than absolute sense... I'm using Gentoo here, so my packages are usually bleeding-edge). So I know it's a lame solution, but you just may have to wait for newer debs.

On a more encouraging note, Wine 0.9.6 was released on the 20th, so you should give that a shot.

[hszFS]

mindpirate, on Jan 25 2006, 05:19 PM, said:

Although you're using the Debian unstable repos, it's likely that at least one of your dependencies is a bit stale ("stale" in a relative rather than absolute sense... I'm using Gentoo here, so my packages are usually bleeding-edge). So I know it's a lame solution, but you just may have to wait for newer debs.

No, this is very unlikely. I was using the latest packages all the time. And now I have tried again with the latest updates, wine 0.9.8 and eMule 0.47a and the problem still persists.

I think it's very obvious that this is caused by wine/eMule, of course it could be some broken debian package but wine is still beta and obviously the primary place to search for the source of a problem.

[mindpirate]

Fair enough. It's probably a bug in Wine then, since nobody seems to have this problem with recent eMule builds on Windows and I can't reproduce the behavior on my Gentoo box.

Please consider filing a bug report with Wine's bugzilla:

http://bugs.winehq.org

Instead of providing them a test case, just refer them back to the HOWTO here so they'll know how you've set up eMule under Wine.

[mindpirate]

Three other open bugs on Wine's bugzilla which directly involve eMule or eMule mods:

• Bug 3063 - emule exposes socket/file handle leak
  
• Bug 3246 - eMulePlus clipping crash
  
• Bug 4299 - Text field malfunction in eMule.exe's search tab

Two more I'm aware of which affect eMule:

• Bug 201 - masking problem - image is not transparent
  
• Bug 3320 - strange looking buttons and graphical glitches

Anyone who can confirm these bugs please do so at Wine's bugzilla.

[mindpirate]

Wine 0.9.9 has been released, and Wine bugs 4299, 201, and 3320 mentioned above have been resolved. After testing this with eMule 0.47a I believe eMule on Linux is now ready for general use. I've tweaked some of the instructions in the HOWTO and bumped its revision to 1.0.

As for the remaining two reported Wine bugs mentioned above, please confirm these on Wine's bugzilla if you run into them.

[hszFS]

mindpirate, on Mar 3 2006, 09:45 AM, said:

Wine 0.9.9 has been released, and Wine bugs 4299, 201, and 3320 mentioned above have been resolved. After testing this with eMule 0.47a I believe eMule on Linux is now ready for general use.

Well, unfortunately you are not right. The 100% CPU-usage bug still persists.
I just upgraded my debian to UNSTABLE, the unstable tree contains the most recent packages. And I also upgraded wine to 0.9.9. The problem still persists.
- WTF!?

[mindpirate]

hszFS, on Mar 14 2006, 09:53 AM, said:

mindpirate, on Mar 3 2006, 09:45 AM, said:

Wine 0.9.9 has been released, and Wine bugs 4299, 201, and 3320 mentioned above have been resolved. After testing this with eMule 0.47a I believe eMule on Linux is now ready for general use.

Well, unfortunately you are not right. The 100% CPU-usage bug still persists.
I just upgraded my debian to UNSTABLE, the unstable tree contains the most recent packages. And I also upgraded wine to 0.9.9. The problem still persists.
- WTF!?

I stand behind my statement that it's ready for general use. Your case is clearly an exception because nobody else has experienced or been able to replicate your problem (or if they have nobody is speaking up, something I'd find strange because people love to bitch when stuff is broken--I know I do). Did you file a bug report with the Wine folks as I suggested? The chances of them fixing a problem they don't know about aren't very good.

[mindpirate]

Sometimes bug votes help raise a bug's visibility/urgency with developers. For those of you experiencing any of the bugs mentioned above, please vote for them on Wine's bugzilla. Registration is quick and painless. After registering simply click on "Vote for this bug" for each bug then allocate your initial 20 votes between them. I usually allocate 5 or more votes per bug.

[leuk_he]

Is it possible to create a dmp of a spinning emule by sending it a kill SIGSEGV? and setting CreateCrashDUmp?

IF yes i can get you a version of Morphmod with debug info (or official ....) to help finding where the mule is spinning.

[mindpirate]

leuk_he, on Mar 31 2006, 03:43 AM, said:

Is it possible to create a dmp of a spinning emule by sending it a kill SIGSEGV? and setting CreateCrashDUmp?

IF yes i can get you a version of Morphmod with debug info (or official ....) to help finding where the mule is spinning.

I'm not sure how you'd go about trapping Linux kernel signals in a Wineserver subprocess that's 100% Windows API.

Luckily Wine has its own debugger, Winedbg, which is supposedly compatible with the Windows debugging API.

http://winehq.com/si...-guide/debugger

Maybe you can rig up something with that?

[bart55]

I've got wine-0.9.11and emule 0.47a

after changing the open files limit /etc/security/limits.conf from 1024 to 30000 works great! no crashes at all!