Bug 147527

Summary: SUSE Help Center takes 2+ minutes to load before asking to build search index
Product: [openSUSE] SUSE Linux 10.1 Reporter: Chad Groneman <cgroneman>
Component: OtherAssignee: E-mail List <kde-maintainers>
Status: RESOLVED FIXED QA Contact: E-mail List <qa-bugs>
Severity: Normal    
Priority: P5 - None    
Version: Beta 3   
Target Milestone: ---   
Hardware: x86-64   
OS: Other   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: strace of susehelp taking long time
Log with new fontconfig RPMs installed.
GZIP of the log file with strace -t -f

Description Chad Groneman 2006-02-02 05:13:28 UTC
When SUSE Help center is launched, at least two minutes go by before the dialog appears to create a search index on dual 2.2 GHz x86-64.  Partial output from 'ps aux' after a couple minutes:

1000      5713  0.0  0.2   9972  2340 pts/0    Ss+  21:48   0:00 bash
1000      5758  0.0  0.1   9356  1616 ?        S    21:56   0:00 /bin/bash /usr/bin/susehelp
1000      5759  0.0  0.1  10268  1604 ?        S    21:56   0:00 wget http://localhost/susehelp/ -O /dev/stdout -q
1000      5760  0.0  0.0   3860   760 ?        S    21:56   0:00 grep <title>SUSE Help
1000      5781  0.0  0.2   9968  2336 pts/1    Ss   21:58   0:00 bash
1000      5801  0.0  0.0   3496   892 pts/1    R+   21:58   0:00 ps aux

I wonder if this might be related to bug 144113.
Comment 1 Michael Gross 2006-02-02 12:42:50 UTC
I don't think it is related, because the error seems to be on the Java-side there. On your output, the process is sleeping. Can you profile the start using `time' and create a backtrace, this will show where the process spends its time. Does this always happen or only at the first start?
Comment 2 Chad Groneman 2006-02-02 15:44:23 UTC
What do you mean by backtrace?  I've heard the term before, but have never done one.  How would you like me to run it, exactly?  Like, 'time strace susehelp' if you  mean strace by backtrace.  Also, if we're timing it, when do you want me to quit running it?
Comment 3 Michael Gross 2006-02-02 16:04:40 UTC
I will try to reproduce this and ask for more information in the case that this is not reproducible, you might wait so long ;) 

If you want to lear more about debugging at you might read:
http://en.opensuse.org/Submit_a_bug#Reporting_a_Bug
Comment 4 Chad Groneman 2006-02-02 20:03:15 UTC
(In reply to comment #3)
> I will try to reproduce this and ask for more information in the case that this
> is not reproducible, you might wait so long ;) 
> 
> If you want to lear more about debugging at you might read:
> http://en.opensuse.org/Submit_a_bug#Reporting_a_Bug
> 

I'll go reproduce this after work, but after skimming through the URL and links, I am still confused as to what is meant by "backtrace".  The -debuginfo RPMs are not available from http://ftp.opensuse.org/pub/opensuse/distribution/SL-10.1-OSS-beta3/ , if that's what you mean.  If you mean get an strace, I can do that. 

I'll go ahead and gather what information I can and attach that tonight (Provo Time).  But it would be helpful to me if you can tell me where to get the -debuginfo RPMs, if that's what generates a backtrace.
Comment 5 Chad Groneman 2006-02-03 01:10:50 UTC
Created attachment 66293 [details]
strace of susehelp taking long time

I did a 'tail -f susehelp.log' and watched where it got stuck.  Afterwards I compared that to the logfile, and found the lag in time to be on line 346.  Line 345 had completely written to the log, but line 346 only had 'wait4(-1, ' for almost the entire time.

Output from timing it:
cgroneman@cg2:~> time strace -o susehelp.log susehelp

real    2m26.105s
user    0m0.096s
sys     0m0.064s
cgroneman@cg2:~>
Comment 6 Michael Gross 2006-02-03 11:37:30 UTC
Chad: This is most likely the fontconfig-bug (bug #147292), please update the packages which you can find here:
http://download.opensuse.org/distribution/SL-10.1-OSS-beta3/fixed-rpms

If the problem still exists, please reopen this bug.
Comment 7 Chad Groneman 2006-02-04 02:43:18 UTC
I applied the RPMs, ran suseconfig, and tried it again.  Roughly the same result, but it told me to run "fc-cache -f" as root.  I did that, went to init 3, then init 5 just to clear things out a bit.  I re-ran my tests.  The output:

cgroneman@cg2:~> time strace -o susehelp2.log susehelp
FC: dircache for /home/cgroneman/.fonts is invalid or outdated,
    consider running "fc-cache -f" as root!

real    2m27.012s
user    0m0.104s
sys     0m0.052s
cgroneman@cg2:~> rpm -qa |grep fontconfig
fontconfig-32bit-2.3.93.20060131-1
fontconfig-2.3.93.20060131-1
fontconfig-devel-32bit-2.3.93.20060131-1
fontconfig-debuginfo-2.3.93.20060131-1
fontconfig-devel-2.3.93.20060131-1
cgroneman@cg2:~>

Problem still happens.  I'll attach the trace.  Can you direct me to where I can get the -debuginfo RPM?  I'm a Novell employee, so I can access internal servers (as long as they have routable IP addresses, ie, not dist.suse.de [10.10.0.100]).
Comment 8 Chad Groneman 2006-02-04 02:46:12 UTC
Created attachment 66454 [details]
Log with new fontconfig RPMs installed.

Again, I did a "tail -f" on the logfile I created, and it got stuck in the same spot.
Comment 9 Michael Gross 2006-02-06 12:47:43 UTC
Might still be the fontconfig problem... KDE-Maintainers: Anything known here?
Comment 10 Dirk Mueller 2006-02-06 12:50:15 UTC

*** This bug has been marked as a duplicate of 143715 ***
Comment 11 Dirk Mueller 2006-02-06 12:53:34 UTC
hmm, actually no indication of fontconfig here. 

can you please post strace -T -f susehelp ?

Comment 12 Dirk Mueller 2006-02-06 12:54:10 UTC
actually its strace -t -f susehelp

thanks. 
Comment 13 Chad Groneman 2006-02-07 00:40:08 UTC
Created attachment 66653 [details]
GZIP of the log file with strace -t -f
Comment 14 Stephan Kulow 2006-02-07 07:52:32 UTC
it's trying very hard to reach a local docserver - I thought it shouldn't do this anymore. Additionally you're suffering from a beta3 kernel bug, that makes connections to localhost fail. I wonder why khelpcenter keeps on trying when it gets a CONNREFUSED though
Comment 15 Dirk Mueller 2006-02-07 09:07:25 UTC
this has nothing to do with dcopserver as far as I can see. it tries to connect to 127.0.0.1:80 and gets a connection refused, and then waits for 2 seconds and tries again until it reached the 2 minutes timeout. 

the beta3 kernel bug hurts but shouldn't be the cause of this. 

very strange. 
Comment 16 Dirk Mueller 2006-02-07 09:09:05 UTC
execve("/usr/bin/wget", ["wget", "http://localhost/susehelp/", "-O", "/dev/stdout", "-q"], [/* 85 vars */] <unfinished ...

there is a wget and its waiting for it. ok, easy fix. 
Comment 17 Dirk Mueller 2006-02-07 09:25:36 UTC
fixed for beta5