[SLE] X applications using Xlib widget starts extremely slow
On a freshly new installation of SuSE 10.1 (machine Pentium III 700Mhz), with only "Basic Graphic environment", traditional X application (should be using Xlib widgets) starts extremely slow, but performs reasonably fast after started. reproduce: 1. login with xdm (gdm not installed) and enter window maker environment. WindowMaker is configured as X session manager by writing it in .xsession 2. Click XTerm icon to launch xterm 3. CPU usage goes to 100%, showing X used up most resources, mouse cursor can hardly move. This goes on for one to two minute until xterm window shows up, then everything goes back normal; 4. type 'gftp' in xterm, gftp starts in less then 10 seconds 5. type 'xmessage hello world' in xterm, CPU usage goes up to 100% like described in 3), the system almost no respond, xmessage window pops up in 48 seconds. Other X application written in traditional way (yudit) also launches extremely slow. On other Linux boxes (e..g Gentoo) xmessage window pops up almost immediately, less then 1 second. While on this SuSE 10.1 it takes 48 seconds. What could probably be the cause? Where can I obtain more debug info / log info to help you determine the cause of my problem? Thanks a lot! P.S. This mail's subject isn't precious actually, because I am not sure if it's really only the X application using Xlib widgets are slow, but I get no better way to describe it. -- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
??? wrote:
On other Linux boxes (e..g Gentoo) xmessage window pops up almost immediately, less then 1 second. While on this SuSE 10.1 it takes 48 seconds.
What could probably be the cause? Where can I obtain more debug info / log info to help you determine the cause of my problem?
Have you tried stracing your app on both platforms? Perhaps AppArmor is gobbling up cycles on SUSE ? /Per Jessen, Zürich -- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com
在 2006-07-09日的 13:48 +0200，Per Jessen写道：
??? wrote:
On other Linux boxes (e..g Gentoo) xmessage window pops up almost immediately, less then 1 second. While on this SuSE 10.1 it takes 48 seconds.
What could probably be the cause? Where can I obtain more debug info / log info to help you determine the cause of my problem?
Have you tried stracing your app on both platforms? Perhaps AppArmor is gobbling up cycles on SUSE ?
I don't know how to "strace" (just in case stace is a term, as I am a bad English speaker). About AppArmor, I am afraid I have no knowledge on it but I can try if I can turn it off. A discover: if I login as root, I can launch these applications very fast (as fast as on other Linux boxes), both in WindowMaker and FVWM. So it could be security-related. I'll see if I can find a clue and post results here. Any other suggestions?
/Per Jessen, Zürich
-- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
??? wrote:
I don't know how to "strace" (just in case stace is a term, as I am a bad English speaker).
strace is a command that shows you all the system calls of your application. You start it like this "strace <your app>". And your English is fine.
About AppArmor, I am afraid I have no knowledge on it but I can try if I can turn it off.
You can disable it in YaST2.
A discover: if I login as root, I can launch these applications very fast (as fast as on other Linux boxes), both in WindowMaker and FVWM. So it could be security-related. I'll see if I can find a clue and post results here.
That's interesting, but I can't really imagine why it behaves like that. /Per Jessen, Zürich -- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com
在 2006-07-09日的 21:10 +0200，Per Jessen写道：
??? wrote:
I don't know how to "strace" (just in case stace is a term, as I am a bad English speaker).
strace is a command that shows you all the system calls of your application. You start it like this "strace <your app>". And your English is fine.
Thank you for enlighten me with 'strace'. I tried to strace the X applications and it turns out the reason is because I set locale to zh_CN.GB18030 Observing strace of xmessage(1), this line of message is printed repeatedly, which seems to be suggesting a locale problem. read(3, "\1\0\212\0\0\0\0\0\0\0R\10\0\0\0\0H\343\355\277\33\277"..., 32) = 32 writev(3, [{"1\0\6\0\1\0\r\0", 8}, {"*-*-GB18030-0", 13}, {"\0\0\0", 3}], 3) = 24 read(3, 0xbfc36edc, 32) = -1 EAGAIN (Resource temporarily unavailable) select(4, [3], NULL, NULL, NULL) = 1 (in [3]) Switch to zh_CN.GB2312 and this problem is immediately solved. output of strace is attached to this email. Do you think I should file a bug? IMHO Linux should behave sanely when using zh_CN.GB18030 locale. GB18030 is the new encoding issued as Chinese mainland standard in 1999, while GB2312 is an old standard issued more then 20 years ago. Old standard GB2312 covers only 6,000 to 7,000 mostly-used Chinese ideographs, compare to a total number of 20,000 Chinese ideographs available today on modern computers. It is making good sense for SuSE to be compatible with the new standard (Though I think UTF-8 is much better choice). Strange thing is, another computer which also suffer from this computer actually uses zh_CN.UTF-8 locale (same locale as the other two computer which do not suffer from this problem). See attachements
About AppArmor, I am afraid I have no knowledge on it but I can try if I can turn it off.
You can disable it in YaST2.
A discover: if I login as root, I can launch these applications very fast (as fast as on other Linux boxes), both in WindowMaker and FVWM. So it could be security-related. I'll see if I can find a clue and post results here.
That's interesting, but I can't really imagine why it behaves like that.
/Per Jessen, Zürich
-- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
see attachments 在 2006-07-10一的 10:06 +0800，张韡武写道：
在 2006-07-09日的 21:10 +0200，Per Jessen写道：
??? wrote:
I don't know how to "strace" (just in case stace is a term, as I am a bad English speaker).
strace is a command that shows you all the system calls of your application. You start it like this "strace <your app>". And your English is fine.
Thank you for enlighten me with 'strace'. I tried to strace the X applications and it turns out the reason is because I set locale to zh_CN.GB18030
Observing strace of xmessage(1), this line of message is printed repeatedly, which seems to be suggesting a locale problem.
read(3, "\1\0\212\0\0\0\0\0\0\0R\10\0\0\0\0H\343\355\277\33\277"..., 32) = 32 writev(3, [{"1\0\6\0\1\0\r\0", 8}, {"*-*-GB18030-0", 13}, {"\0\0\0", 3}], 3) = 24 read(3, 0xbfc36edc, 32) = -1 EAGAIN (Resource temporarily unavailable) select(4, [3], NULL, NULL, NULL) = 1 (in [3])
Switch to zh_CN.GB2312 and this problem is immediately solved.
output of strace is attached to this email.
Do you think I should file a bug? IMHO Linux should behave sanely when using zh_CN.GB18030 locale. GB18030 is the new encoding issued as Chinese mainland standard in 1999, while GB2312 is an old standard issued more then 20 years ago. Old standard GB2312 covers only 6,000 to 7,000 mostly-used Chinese ideographs, compare to a total number of 20,000 Chinese ideographs available today on modern computers. It is making good sense for SuSE to be compatible with the new standard (Though I think UTF-8 is much better choice).
Strange thing is, another computer which also suffer from this computer actually uses zh_CN.UTF-8 locale (same locale as the other two computer which do not suffer from this problem). See attachements
About AppArmor, I am afraid I have no knowledge on it but I can try if I can turn it off.
You can disable it in YaST2.
A discover: if I login as root, I can launch these applications very fast (as fast as on other Linux boxes), both in WindowMaker and FVWM. So it could be security-related. I'll see if I can find a clue and post results here.
That's interesting, but I can't really imagine why it behaves like that.
/Per Jessen, Zürich
-- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
??? wrote:
Thank you for enlighten me with 'strace'. I tried to strace the X applications and it turns out the reason is because I set locale to zh_CN.GB18030
Observing strace of xmessage(1), this line of message is printed repeatedly, which seems to be suggesting a locale problem.
read(3, "\1\0\212\0\0\0\0\0\0\0R\10\0\0\0\0H\343\355\277\33\277"..., 32) = 32 writev(3, [{"1\0\6\0\1\0\r\0", 8}, {"*-*-GB18030-0", 13}, {"\0\0\0", 3}], 3) = 24 read(3, 0xbfc36edc, 32) = -1 EAGAIN (Resource temporarily unavailable) select(4, [3], NULL, NULL, NULL) = 1 (in [3])
Switch to zh_CN.GB2312 and this problem is immediately solved.
Excellent.
Do you think I should file a bug? IMHO Linux should behave sanely when using zh_CN.GB18030 locale. GB18030 is the new encoding issued as Chinese mainland standard in 1999, while GB2312 is an old standard issued more then 20 years ago. Old standard GB2312 covers only 6,000 to 7,000 mostly-used Chinese ideographs, compare to a total number of 20,000 Chinese ideographs available today on modern computers. It is making good sense for SuSE to be compatible with the new standard (Though I think UTF-8 is much better choice).
It sounds like it's worth filing a bug, yes. I am not at all familiar with chinese encodings/locales, but if GB18030 is the new standard, SUSE will certainly want to be able to work with it. /Per Jessen, Zürich -- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com
在 2006-07-10一的 08:19 +0200，Per Jessen写道：
??? wrote:
Thank you for enlighten me with 'strace'. I tried to strace the X applications and it turns out the reason is because I set locale to zh_CN.GB18030
Observing strace of xmessage(1), this line of message is printed repeatedly, which seems to be suggesting a locale problem.
read(3, "\1\0\212\0\0\0\0\0\0\0R\10\0\0\0\0H\343\355\277\33\277"..., 32) = 32 writev(3, [{"1\0\6\0\1\0\r\0", 8}, {"*-*-GB18030-0", 13}, {"\0\0\0", 3}], 3) = 24 read(3, 0xbfc36edc, 32) = -1 EAGAIN (Resource temporarily unavailable) select(4, [3], NULL, NULL, NULL) = 1 (in [3])
Switch to zh_CN.GB2312 and this problem is immediately solved.
Excellent.
Do you think I should file a bug? IMHO Linux should behave sanely when using zh_CN.GB18030 locale. GB18030 is the new encoding issued as Chinese mainland standard in 1999, while GB2312 is an old standard issued more then 20 years ago. Old standard GB2312 covers only 6,000 to 7,000 mostly-used Chinese ideographs, compare to a total number of 20,000 Chinese ideographs available today on modern computers. It is making good sense for SuSE to be compatible with the new standard (Though I think UTF-8 is much better choice).
It sounds like it's worth filing a bug, yes. I am not at all familiar with chinese encodings/locales, but if GB18030 is the new standard, SUSE will certainly want to be able to work with it.
/Per Jessen, Zürich
Bug filed. https://bugzilla.novell.com/show_bug.cgi?id=191487 -- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
在 2006-07-09日的 13:48 +0200，Per Jessen写道：
??? wrote:
On other Linux boxes (e..g Gentoo) xmessage window pops up almost immediately, less then 1 second. While on this SuSE 10.1 it takes 48 seconds.
What could probably be the cause? Where can I obtain more debug info / log info to help you determine the cause of my problem?
Have you tried stracing your app on both platforms? Perhaps AppArmor is gobbling up cycles on SUSE ?
I cannot turn off AppArmor because it is not installed (I think it is not installed, because I don't see AppArmor in yast2). Further the firewall is already turned off. I just checked 4 SuSE computers (all in 10.1 version) in the office and found * two of them suffer from this problem (a.k.a traditional X applications like xterm starts very slow). One computer has this problem both as root and as normal user (Pentium 2.4GHz, Matrox G450 video) the other one only have problem as normal user (Pentium 700MHz, 815 on-board video); * two other computers do not suffer from this problem, both as root user and as normal user (both > 2.GHz CPU, one with ATI card and the other with onboard 845 card)
/Per Jessen, Zürich
-- 锐业软服（国内业务） http://www.realss.cn Real SoftService http://www.realss.com 销售咨询(Sales Department): 0086 592 20 99987 (Chinese, German, English) 国际业务(International Sales): 0086 10 8460 6011 (German and English) 联系：厦门大学科技园，嘉庚二号楼6楼 邮政：厦门大学2312号信箱（邮编361005）
participants (2)
-
Per Jessen -
张韡武