You can subscribe to this list here.
| 2002 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(2) |
Jun
|
Jul
(3) |
Aug
|
Sep
|
Oct
(1) |
Nov
|
Dec
(1) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2003 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
(2) |
Sep
|
Oct
|
Nov
|
Dec
|
| 2011 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
(1) |
Dec
|
| 2012 |
Jan
(1) |
Feb
(1) |
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(1) |
Oct
(1) |
Nov
|
Dec
(1) |
| 2013 |
Jan
|
Feb
|
Mar
(2) |
Apr
|
May
|
Jun
|
Jul
|
Aug
(1) |
Sep
|
Oct
|
Nov
(3) |
Dec
|
| 2014 |
Jan
|
Feb
|
Mar
|
Apr
(12) |
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
(6) |
| 2015 |
Jan
(13) |
Feb
|
Mar
|
Apr
|
May
(3) |
Jun
|
Jul
|
Aug
|
Sep
(2) |
Oct
|
Nov
|
Dec
|
| 2016 |
Jan
|
Feb
|
Mar
|
Apr
(3) |
May
|
Jun
|
Jul
|
Aug
(3) |
Sep
|
Oct
|
Nov
(3) |
Dec
(9) |
| 2017 |
Jan
|
Feb
|
Mar
|
Apr
(3) |
May
(2) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
(1) |
Dec
|
| 2020 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
(3) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2021 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
(3) |
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2022 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(3) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2023 |
Jan
|
Feb
(4) |
Mar
|
Apr
|
May
(5) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2024 |
Jan
|
Feb
|
Mar
(1) |
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2026 |
Jan
(2) |
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: Mark H. <ma...@xw...> - 2026-01-13 22:11:08
|
With so few changes to xosview 1.x, it may be better to adopt some kind 'rolling' release. eg. asking distributions to use the Git "master". I'm sure others have experience here. The problems it aims to improve: * small patches typically end up waiting for a release * encourage distributions to upstream patches early The release process is not a huge burden, but after any activity triggers a waiting game -- as its the period when more patches are most likely. Distributions may still require version numbers for packages, in which case it could be: 1.26.x where "x" is the number of patches since 1.26.1 Any thoughts or improvements to this proposal? Thanks -- Mark |
|
From: Mark H. <ma...@xw...> - 2026-01-13 21:55:13
|
I just tagged the few patches in the Git repo as a new release: https://www.pogo.org.uk/~mark/xosview/ https://www.pogo.org.uk/~mark/xosview/releases/xosview-1.25.tar.gz These are minor fixes; with thanks to Dr. Werner Fink, Kartik Mistry and Andreas K. Hüttel. -- Mark |
|
From: Mark H. <ma...@xw...> - 2024-03-25 16:59:59
|
One of my favoured distributions at the moment, but it was missing an xosview package so I submitted one [*]. Today it was merged to "testing". If anyone is an Alpine enthusiast and interested to assist in maintaining the package, you would be most welcome. So now with the following in /etc/apk/repositories on v3.19: @testing http://dl-cdn.alpinelinux.org/alpine/edge/testing I was able to do: # apk add xosview@testing [*] https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/61708 -- Mark |
|
From: Dale R. W. <wo...@al...> - 2023-05-24 18:11:33
|
mik...@co... writes: > There kinda is. I don't believe it is in the man page but it is in > the Xdefaults file: > > xosview*samplesPerSec: 10 > > So, this makes things a but more complicated (which is why it is not > documented :). Lemme try and explain... "E.g. the period for updating the LOAD display is loadPriority/samplesPerSec seconds." It seems to work that way for me, on the old version I've got. Thanks! Dale |
|
From: <mik...@co...> - 2023-05-23 20:17:16
|
>>>>> " " == Dale R Worley <wo...@al...> writes:
> mik...@co... writes:
>> Search the man page for 'priority'. For example:
>> xosview*prefixPriority: \fIvalue\fP
> Thanks!
> Too bad there isn't a master resource that sets the default for
> all the Priority resources. But this does work for me:
There kinda is. I don't believe it is in the man page but it is in
the Xdefaults file:
xosview*samplesPerSec: 10
So, this makes things a but more complicated (which is why it is not
documented :). Lemme try and explain...
The loadPriority is normally 20. Which would cause it to update once
every 2 seconds if samplesPerSec is the default of 10. If samples per
second were instead set to 5. Then I *think* (untested) the loadmeter
(set at 20) would then update every 4 seconds. Test this out as it has been
decades since I've tried it :).
Mike
|
|
From: Dale R. W. <wo...@al...> - 2023-05-23 18:08:28
|
mik...@co... writes:
> Search the man page for 'priority'. For example:
>
> xosview*prefixPriority: \fIvalue\fP
Thanks!
Too bad there isn't a master resource that sets the default for all the
Priority resources. But this does work for me:
xosview \
-xrm "xosview*loadPriority: $P" -xrm "xosview*cpuPriority: $P" \
-xrm "xosview*memPriority: $P" -xrm "xosview*swapPriority: $P" \
-xrm "xosview*pagePriority: $P" -xrm "xosview*netPriority: $P" \
-xrm "xosview*batteryPriority: $P" -xrm "xosview*wirelessPriority: $P" \
-xrm "xosview*diskPriority: $P"
Dale
|
|
From: <mik...@co...> - 2023-05-22 17:02:52
|
>>>>> " " == Dale R Worley <wo...@al...> writes:
> Watching xosview, it's clear that its internal cycle is quite
> short, perhaps as low as a tenth of a second. Unfortunately,
> I've got a situation where xosview's CPU consumption is too
> high. Is there any way to increase the period of its inside
> loop? I've searched in the manual page but not found anything
> which claims to control its speed of execution. Though the
> manual page is over 20 pages of text and I may have overlooked
> something.
Search the man page for 'priority'. For example:
xosview*prefixPriority: \fIvalue\fP
.RS
This number (which must be an integer >= 1) sets the number of tenths of
a second that the meter waits between updates. A value of 1 has xosview
update the meter 10 times per second (the fastest). A value of 600 would
cause xosview to update the meter once a minute.
.RE
Each meter has a different setting. For instance the cpu meter is set
by default to:
xosview*cpuPriority: 2
Which causes it to update five times per second. If you change this to
10 then the cpu meter would update once per second.
Mike
|
|
From: Dale R. W. <wo...@al...> - 2023-05-22 01:24:51
|
Watching xosview, it's clear that its internal cycle is quite short, perhaps as low as a tenth of a second. Unfortunately, I've got a situation where xosview's CPU consumption is too high. Is there any way to increase the period of its inside loop? I've searched in the manual page but not found anything which claims to control its speed of execution. Though the manual page is over 20 pages of text and I may have overlooked something. Thanks, Dale |
|
From: John P. A. G. <gla...@ph...> - 2023-02-02 16:17:49
|
Hi Mark! On Thu, 2023-02-02 at 16:07 +0000, Mark Hills wrote: > > > Before I tag a release, is Debian running with any > > > local patches at the moment? > > > > There are no local patches at the moment: > > > > > https://salsa.debian.org/debian/xosview/-/tree/master/debian > > Ok great, I did that. Please can you check the web page for accuracy and > test the .tar.gz? > > http://www.pogo.org.uk/~mark/xosview/ Looks good. Building a test package for both Debian and openSUSE each worked without any issues. Also did a quick test run. Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer `. `' Physicist `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913 |
|
From: Mark H. <ma...@xw...> - 2023-02-02 16:07:55
|
On Thu, 2 Feb 2023, John Paul Adrian Glaubitz wrote: > Hi Mark! > > On Thu, 2023-02-02 at 15:55 +0000, Mark Hills wrote: > > On Thu, 2 Feb 2023, John Paul Adrian Glaubitz wrote: > > > > > I was wondering whether there are any plans for a new minor release of > > > xosview in the near future? [...] > > Before I tag a release, is Debian running with any > > local patches at the moment? > > There are no local patches at the moment: > > > https://salsa.debian.org/debian/xosview/-/tree/master/debian Ok great, I did that. Please can you check the web page for accuracy and test the .tar.gz? http://www.pogo.org.uk/~mark/xosview/ -- Mark |
|
From: John P. A. G. <gla...@ph...> - 2023-02-02 16:00:40
|
Hi Mark! On Thu, 2023-02-02 at 15:55 +0000, Mark Hills wrote: > On Thu, 2 Feb 2023, John Paul Adrian Glaubitz wrote: > > > I was wondering whether there are any plans for a new minor release of > > xosview in the near future? > > Can certainly be done. Great! > Before I tag a release, is Debian running with any > local patches at the moment? There are no local patches at the moment: > https://salsa.debian.org/debian/xosview/-/tree/master/debian These would be in the debian/patches folder. Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer `. `' Physicist `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913 |
|
From: Mark H. <ma...@xw...> - 2023-02-02 15:55:32
|
On Thu, 2 Feb 2023, John Paul Adrian Glaubitz wrote: > I was wondering whether there are any plans for a new minor release of > xosview in the near future? Can certainly be done. Before I tag a release, is Debian running with any local patches at the moment? > I'm asking since there have been a number of fixes in the code since the > last release [1] and since a new Debian stable releases is around the > corner, it would be nice to have the next version of xosview included in > the coming Debian release. > > There is still enough time to upload the new release to Debian since the soft > freeze will only start in ten days (2023-02-12) [2]. > > Thanks, > Adrian > > > [1] https://github.com/hills/xosview/commits/master > > [2] https://release.debian.org/bookworm/freeze_policy.html > > -- Mark |
|
From: George R G. <gr...@ya...> - 2022-05-31 08:13:28
|
Mike, The build has succeeded so THANKS for your help.- I was looking for the command line switch to specify an interval regarding updating the screen. Xosview2 appears to be updating the screen very quickly... showing cpu time in particular as shown by "xosview2 +cpus". I have tried (perhaps in error) the Xresource "xosview*prefixPriority: 200" which I put in my .Xdefaults file and pointed xrdb -merge to this file. Any thoughts? I'm using WindowMaker and not KDE... It appears to be faster and "smaller" than KDE. THANKS again for responding to this bug report. Best regards, George... |
|
From: <mik...@co...> - 2022-05-14 00:39:49
|
>>>>> " " == George R Goffe via Xosview-users <xos...@li...> writes:
> Hi, I'm trying to build xosview2 from the git repo at
> https://git.code.sf.net/p/xosview/git but am experiencing a
> problem. The code I have cloned seems to think it's "already up
> to date".
'''
Looks like the experimental XSM (X session management) is broken. Try
running configure with the --without-xsm switch. Or just don't use the
--with-xsm switch.
I'll look into updating the xsm support. However, I
never found that kde implemented this according to the spec. And the whole
wayland mess is not making things more promising :).
Mike
|
|
From: George R G. <gr...@ya...> - 2022-05-11 07:16:19
|
Hi, I'm trying to build xosview2 from the git repo at https://git.code.sf.net/p/xosview/git but am experiencing a problem. The code I have cloned seems to think it's "already up to date". Here's what I'm seeing... Am I doing something wrong? Best regards and STAY SAFE! George... c++ -Wall -Wextra -O2 -I/usr/X11R6/include -I/usr/include/freetype2 -Wp,-MD,.xosview.o.d -I/tools/xosview/xosview2 -I/tools/xosview/xosview2/cmeter -I/tools/xosview/xosview2/linux -DHAVE_CONFIG_H -c xosview.cc -o xosview.o xosview.cc: In member function ‘void XOSView::loadConfiguration(std::vector<std::__cxx11::basic_string<char> >&)’: xosview.cc:318:21: error: ‘const class util::CLOpt’ has no member named ‘eraseFrom’ 318 | opt.eraseFrom(argv); | ^~~~~~~~~ make: *** [Makefile.config:42: xosview.o] Error 1 |
|
From: Gösta L. <gos...@fo...> - 2021-06-17 15:28:57
|
Thanks for the workaround and the segfault fix. Now there's only the BUG note. Turned out that I had enabled core temperatures in ~/.Xresources on the Precision T5810. That was more for fun and I had totally forgot about it when I copied the file to the Precision 5820. I can certainly live without the temperatures so disabling them solves it for me just fine.
________________________________
Från: Russell Reed <Rus...@ac...>
Skickat: den 17 juni 2021 15:21:31
Till: Gösta Ljungdahl; xos...@li...
Ämne: RE: xosview 1.21 and 1.23 segfault
It looks like the reason you’re getting a segfault rather than more graceful handling of the issue is due to a missing return statement in the block on lines 55-58 of linux/coretemp.cc
if ( _cpus.empty() ) { // should not happen at this point
std::cerr << "BUG: Could not determine sysfs file(s) for coretemp." << std::endl;
parent_->done(1);
}
It should look like this instead:
if ( _cpus.empty() ) { // should not happen at this point
std::cerr << "BUG: Could not determine sysfs file(s) for coretemp." << std::endl;
parent_->done(1);
return;
}
That won’t make the temperature meter work, though.
Assuming that you don’t care about the temperature meter, the easiest way to workaround this would be to turn the temperature meter off. It defaults to off, so apparently your configuration has turned it on for this machine.
Add the following line to the .Xdefaults file in the home directory of the user you’re running xosview as:
xosview*coretemp: False
If the .Xdefaults file doesn’t exist, just create it with this one line.
If it does already exist and it already has a line for xosview*coretemp (presumably with a value of True), change that line to False instead of adding another line.
From: Gösta Ljungdahl <gos...@fo...>
Sent: Thursday, June 17, 2021 5:47 AM
To: xos...@li...
Subject: [Xosview-users] xosview 1.21 and 1.23 segfault
I can run xosview v 1.21 and also 1.23 on CentosOS 7.9.2009 compiled using system gcc 4.8.5 on a Dell Precision T5810, Intel Xeon E5-1650 v4 @ 3.60GHz, 72 GB RAM, Radeon Pro WX 7100 driver amdgpu.
But I get a segfault on Dell Precision 5820, Intel Xeon W-2125 @ 4.00GHz, 64 GB RAM, nVidia Quadro P2000 driver nvidia. Same kernel on both systems i.e. 3.10.0-1160.25.1.el7.x86_64.
The error message in the terminal window is this:
BUG: Could not determine sysfs file(s) for coretemp.
The same thing happens with version 1.23 and this is what I get from dmesg:
xosview[3701]: segfault at 0 ip 00007f505c654f78 sp 00007ffd2258ede0 error 4 in 1ibstdc++.so.6.0.19[7f505c596000+e9000]
systemd journal gives me (truncated lines):
jun 16 17:48:02 localhost.localdomain kernel: xosview[3701] : segfault at 0 ip 00007f505c654f78 sp 00
jun 16 17:48:02 localhost.localdomain abrt-hook-ccpp[3702]: Process 3701 (xosview) of user 2627 kill
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Executable '/usr/local/xosview1.23/bin/xosv
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : 'post-create' on '/var/spool/abrt/ccpp-2021
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Deleting problem directory '/var/spool/abrt
As you can see abrt deletes the dump but I've found out how to avoid that but haven't done it yet.
By using bc I've established the location of the segfault:
7f505c654f78-7f505c596000=bef78
objdump --demangle -d /usr/lib64/libstdc++.so.6
...
bef78: 48 8b 06 mov (%rsi),%rax
...
but that's as far as I got.
For some reason the code doesn't get it right on the Xeon W-2125 processor. This processor sits in an off-line computer so getting info out of it boils down to printouts and the OCR-scanning those.
Any help in troubleshooting/fixing is greatly appreciated.
Kind regards,
Gösta Ljungdahl
***************************************************************************
The information contained in this communication is confidential, is
intended only for the use of the recipient named above, and may be legally
privileged.
If the reader of this message is not the intended recipient, you are
hereby notified that any dissemination, distribution or copying of this
communication is strictly prohibited.
If you have received this communication in error, please resend this
communication to the sender and delete the original message or any copy
of it from your computer system.
Thank You.
****************************************************************************
|
|
From: Russell R. <Rus...@ac...> - 2021-06-17 13:37:15
|
It looks like the reason you're getting a segfault rather than more graceful handling of the issue is due to a missing return statement in the block on lines 55-58 of linux/coretemp.cc
if ( _cpus.empty() ) { // should not happen at this point
std::cerr << "BUG: Could not determine sysfs file(s) for coretemp." << std::endl;
parent_->done(1);
}
It should look like this instead:
if ( _cpus.empty() ) { // should not happen at this point
std::cerr << "BUG: Could not determine sysfs file(s) for coretemp." << std::endl;
parent_->done(1);
return;
}
That won't make the temperature meter work, though.
Assuming that you don't care about the temperature meter, the easiest way to workaround this would be to turn the temperature meter off. It defaults to off, so apparently your configuration has turned it on for this machine.
Add the following line to the .Xdefaults file in the home directory of the user you're running xosview as:
xosview*coretemp: False
If the .Xdefaults file doesn't exist, just create it with this one line.
If it does already exist and it already has a line for xosview*coretemp (presumably with a value of True), change that line to False instead of adding another line.
From: Gösta Ljungdahl <gos...@fo...>
Sent: Thursday, June 17, 2021 5:47 AM
To: xos...@li...
Subject: [Xosview-users] xosview 1.21 and 1.23 segfault
I can run xosview v 1.21 and also 1.23 on CentosOS 7.9.2009 compiled using system gcc 4.8.5 on a Dell Precision T5810, Intel Xeon E5-1650 v4 @ 3.60GHz, 72 GB RAM, Radeon Pro WX 7100 driver amdgpu.
But I get a segfault on Dell Precision 5820, Intel Xeon W-2125 @ 4.00GHz, 64 GB RAM, nVidia Quadro P2000 driver nvidia. Same kernel on both systems i.e. 3.10.0-1160.25.1.el7.x86_64.
The error message in the terminal window is this:
BUG: Could not determine sysfs file(s) for coretemp.
The same thing happens with version 1.23 and this is what I get from dmesg:
xosview[3701]: segfault at 0 ip 00007f505c654f78 sp 00007ffd2258ede0 error 4 in 1ibstdc++.so.6.0.19[7f505c596000+e9000]
systemd journal gives me (truncated lines):
jun 16 17:48:02 localhost.localdomain kernel: xosview[3701] : segfault at 0 ip 00007f505c654f78 sp 00
jun 16 17:48:02 localhost.localdomain abrt-hook-ccpp[3702]: Process 3701 (xosview) of user 2627 kill
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Executable '/usr/local/xosview1.23/bin/xosv
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : 'post-create' on '/var/spool/abrt/ccpp-2021
jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Deleting problem directory '/var/spool/abrt
As you can see abrt deletes the dump but I've found out how to avoid that but haven't done it yet.
By using bc I've established the location of the segfault:
7f505c654f78-7f505c596000=bef78
objdump --demangle -d /usr/lib64/libstdc++.so.6
...
bef78: 48 8b 06 mov (%rsi),%rax
...
but that's as far as I got.
For some reason the code doesn't get it right on the Xeon W-2125 processor. This processor sits in an off-line computer so getting info out of it boils down to printouts and the OCR-scanning those.
Any help in troubleshooting/fixing is greatly appreciated.
Kind regards,
Gösta Ljungdahl
***************************************************************************
The information contained in this communication is confidential, is
intended only for the use of the recipient named above, and may be legally
privileged.
If the reader of this message is not the intended recipient, you are
hereby notified that any dissemination, distribution or copying of this
communication is strictly prohibited.
If you have received this communication in error, please resend this
communication to the sender and delete the original message or any copy
of it from your computer system.
Thank You.
****************************************************************************
|
|
From: Gösta L. <gos...@fo...> - 2021-06-17 11:02:37
|
I can run xosview v 1.21 and also 1.23 on CentosOS 7.9.2009 compiled using system gcc 4.8.5 on a Dell Precision T5810, Intel Xeon E5-1650 v4 @ 3.60GHz, 72 GB RAM, Radeon Pro WX 7100 driver amdgpu. But I get a segfault on Dell Precision 5820, Intel Xeon W-2125 @ 4.00GHz, 64 GB RAM, nVidia Quadro P2000 driver nvidia. Same kernel on both systems i.e. 3.10.0-1160.25.1.el7.x86_64. The error message in the terminal window is this: BUG: Could not determine sysfs file(s) for coretemp. The same thing happens with version 1.23 and this is what I get from dmesg: xosview[3701]: segfault at 0 ip 00007f505c654f78 sp 00007ffd2258ede0 error 4 in 1ibstdc++.so.6.0.19[7f505c596000+e9000] systemd journal gives me (truncated lines): jun 16 17:48:02 localhost.localdomain kernel: xosview[3701] : segfault at 0 ip 00007f505c654f78 sp 00 jun 16 17:48:02 localhost.localdomain abrt-hook-ccpp[3702]: Process 3701 (xosview) of user 2627 kill jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Executable '/usr/local/xosview1.23/bin/xosv jun 16 17:48:03 localhost.localdomain abrt-server[3703] : 'post-create' on '/var/spool/abrt/ccpp-2021 jun 16 17:48:03 localhost.localdomain abrt-server[3703] : Deleting problem directory '/var/spool/abrt As you can see abrt deletes the dump but I've found out how to avoid that but haven't done it yet. By using bc I've established the location of the segfault: 7f505c654f78-7f505c596000=bef78 objdump --demangle -d /usr/lib64/libstdc++.so.6 ... bef78: 48 8b 06 mov (%rsi),%rax ... but that's as far as I got. For some reason the code doesn't get it right on the Xeon W-2125 processor. This processor sits in an off-line computer so getting info out of it boils down to printouts and the OCR-scanning those. Any help in troubleshooting/fixing is greatly appreciated. Kind regards, Gösta Ljungdahl |
|
From: Kyle T. <kyl...@gm...> - 2020-07-06 16:55:47
|
On 7/6/20 9:31 AM, Russell Reed via Xosview-users wrote:
> Also take a look at that fourth field in /proc/loadavg. That's the number of threads running or ready to run and the total number of threads in the system. When your load average is high, the number of threads currently running should also be high, when it's working correctly.
Another possible cause I recall is related to hung IO.
Check the output of ‘ps aux’. If any process is in the ‘D’ state,
then it is in a deep uninterruptible state -- usually hung IO. Such a
process increments the load average, even though it isn’t really doing
anything. This is a common problem on NFS setups when the NFS source
goes down.
--
[*] Kyle Terrien
Linux Troubleshooter
I help you get to the *root* of your problems.
< kyl...@gm...
* https://www.linkedin.com/in/kyleterrien/
* https://klipkyle.gitlab.io/
* Telegram: t.me/klipkyle
|
|
From: Russell R. <Rus...@ac...> - 2020-07-06 16:47:10
|
Xosview on Linux just gets the load average from the first field in the /proc/loadavg special file. Example (on my system, the current load average is 0.88): $ cat /proc/loadavg 0.88 0.22 0.07 4/318 12628 If you look at /proc/loadavg and that value is too high, then it is a kernel issue. If the value in /proc/loadavg looks fine, but xosview is off, then it's probably an issue with how xosview processes the value. Also take a look at that fourth field in /proc/loadavg. That's the number of threads running or ready to run and the total number of threads in the system. When your load average is high, the number of threads currently running should also be high, when it's working correctly. It looks like there has been one change in the kernel this year to how the load average is calculated. This issue might be related to that change. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/kernel/sched/loadavg.c?h=v5.8-rc4&id=ebc0f83c78a2d26384401ecf2d2fa48063c0ee27 -----Original Message----- From: George R Goffe via Xosview-users <xos...@li...> Sent: Monday, July 6, 2020 10:49 AM To: xos...@li... Subject: [Xosview-users] xosview and xosview2 report HUGE load averages Hi, I'm running a Fedora Core 33 (Rawhide) system and have noticed that several of the tools that refer to load average are reporting HUGE numbers. Top and htop as well as the two versions of xosview are affected. Current kernel is 5.8.0-0.rc3.20200630git7c30b859a947.1.fc33.x86_64. I have reported this to the RedHat bugzilla, bug-id: 1854158 – htop reports load average too large Can anyone help me with this situation? Please? George... 1854158 – htop reports load average too large _______________________________________________ Xosview-users mailing list Xos...@li... https://lists.sourceforge.net/lists/listinfo/xosview-users *************************************************************************** The information contained in this communication is confidential, is intended only for the use of the recipient named above, and may be legally privileged. If the reader of this message is not the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please resend this communication to the sender and delete the original message or any copy of it from your computer system. Thank You. **************************************************************************** |
|
From: George R G. <gr...@ya...> - 2020-07-06 15:49:05
|
Hi,
I'm running a Fedora Core 33 (Rawhide) system and have noticed that several of the tools that refer to load average are reporting HUGE numbers. Top and htop as well as the two versions of xosview are affected. Current kernel is 5.8.0-0.rc3.20200630git7c30b859a947.1.fc33.x86_64.
I have reported this to the RedHat bugzilla, bug-id: 1854158 – htop reports load average too large
Can anyone help me with this situation? Please?
George...
1854158 – htop reports load average too large
|
|
From: Mark H. <ma...@po...> - 2017-11-05 21:51:20
|
Just a small update to xosview 1.x: http://www.pogo.org.uk/~mark/xosview/ http://www.pogo.org.uk/~mark/xosview/releases/xosview-1.20.tar.gz The repository is hosted on Github: https://github.com/hills/xosview The most significant change is a fix to a bug introduced in 1.17, where xosview could crash if network devices are removed on a running system (eg. VM bridge, wireless) -- Mark |
|
From: L A W. <xo...@tl...> - 2017-05-28 03:53:45
|
Sorry for the delay in getting back -- I don't read the list
mail as regularly as personal mail -- i.e. if you respond to me, it
"help" if you include me on the To or CC line... ;-).
That may be partly due to the mail I sent to you & list being
separated in time due to it taking a while to get the list mail
to go through... *sigh*...
mik...@co... wrote:
>> " " == L A Walsh writes:
>>
>
> > The flexibility isn't really at issue. In fact, to
> > maintain compatibility, any initial splitting off of the config
> > would likely start by duplicating parts of the Xrm.
>
> If X resources were replaced, then the sky is the limit. No reason
> why any existing interface would need to be kept. All that is needed
> is some clean way for users to specify options and the code to use
> them. The key thing is that the interface be clean and easy to use
> for both. :)
>
----
Yeah, well that's the 'rub'.. one person's definition of clean
OR easy, might differ from another's. But yeah, that's the goal!
> [...]
>
> > But however & wherever the configs are,
> > currently, they are mixed in with display parameters specific
> > to X. How they are mixed in and the flexibility of that system
> > doesn't seem to be an essential feature of an OS-monitor
> > program.
>
> Not sure I'm following you exactly here.
---
Font info, window sizing & placement -- those are display issues.
Meter ranges, device names, Meter-sub-fields, those are more meter-config
issues.
> Are you talking about the
> distinction between what to display and how to display it? As I see
> things the configurable options can be broken down into two basic
> categories:
>
> 1) What information to display: (cpu, memory, swap, etc).
>
> 2) How to display it: (color, type of meter, maximum values, fonts,
> etc)
>
Yeah, I think we are pretty much on the same page, though not all
types of meters make sense for the various information values. While
an instantaneous reading of a disk or network speed would work, one almost
always wants some history.
> I suppose #2 could be broken down even further into pure "graphics"
> type options (color, font, etc) and everything else. Regardless of
> how one would subdivide things, I'm not sure of the benefit unless the
> data collecting code were split from the display code in some way.
> Which is what I think you are driving at.
>
----
Display could be buffered or logged to disk or data could be sent
to another system for collection/processing -- sent via RPC or UDP (I'm
thinking, eventually, it might be useful to have a stateless mechanism
(vs. TCP/RPC) since if people bring up monitoring for a system, from a
portable or handheld, it would be good to be able to pause the display
as a device is suspended with the data being buffered to disk (NOTE: I'm
brainstorming, not working on such!)...but when they reconnect, it would
be nice to allow the user to scroll sideways to look back in time,
perhaps. or another system via an RPC or UDP mechanism.
In my last working version (from 2+ years ago), I had the recording
of the data separate from the graph size. I used a buffer size the width
of my screen so I could stretch the graph to the full window and have it
display 2560 time-units back, but usually I had the window no more than
600 wide or so -- but could still stretch it or shrink it and it would
display up to the max-units saved.
That version also filled the full graph width w/the first measurement then
used the movearea & draw to slowly fill in the entire width (at no time
needing to redraw the entire line unless it needed to refresh) -- very low
bandwidth & cpu usage.
> Doing such a split would probably best be done by first having a
> second viable display system to use first. ...
>
> > I'm not a web expert, but it seems that one might have a
> > suitable graphics framework using an HTML5-based canvas (?).
> ...
>
> I have thought about this from time to time. As I see it,
> displaying system related information as a "web app" would require a
> few things:
>
> 1) Some code which runs on the actual hardware being monitored.
> Because of security concerns this code would not be loaded from a
> http server. Probably it would have to be some sort of "stats"
> daemon running on the hardware already.
>
----
Agreed -- as in you wanting stats from your server farm.
> 2) Some network interface for this daemon/server to talk to clients
> with. Since the client will be a web browser this interface
> would probably best be an http server. Which would mean the code
> in #1 would either need to implement an http server or be part of
> an existing one.
>
> 2) Some way to load "client code" into the web browser (html,
> javascript, etc). I'm not that familiar with html5 canvases.
> But I think this would involve html and probably javascript.
>
> 3) Some way to get the information from #1 into the #2. There are
> some http/html methods I'm aware of to do this (SSE, etc). So,
> it is probably possible to open a socket that stays open for a
> while between a server and an http client. But how long it stays
> open may depend a lot on both the server and client browser. It
> would probably also be best to have this connection use UDP
> packets. But I'm not aware of any web standards in that regard.
>
-----
Yeah.. not sure if web server if the best for remote clients
using a handheld or not. I just don't know enough about web stuff
to know. Still if a remote platform ran a linux (like android), it might
run an X client, that might display data retrieved via some mechanism...
but thus my preference to try to separate display and config, maybe even
transport, to allow for flexibility. If we stick to what we can imagine
today, it might be a bear (more of one than it already is) to rearchitect.
>
> One could probably do some sort of proof of concept of the above by
> using something like cherrypy to server up a simple http server which
> would send out cpu stats to a html page with an html5 canvas. I'm
> not sure how good or bad the performance of something like this would
> be. If the network connection is reasonable, X itself works ok
> remotely. So, in theory it should be possible. Just not sure how
> much of an unknown a web browser would be.
>
ditto...
>
> > FWIW, I have no immediate plans to try to port
> > it elsewhere, just thinking about modular & OO design (maybe
> > from reading Stroustrup's 2011 version of C++).
>
> It will be nice once c++11 compilers are the default on most
> platforms. Last I checked (about a year ago), c++98 compilers were
> still the default on many platforms. This may change in the next
> year or two. Until then, I am hesitant to use any c++11 or newer
> features since it would require many folks to first install a compiler.
>
---
Well... c++17 compilers are already out (though not by that
name), as are c++14 and c++11. If you aren't careful, those
compiling the product will have to install old compat libs to work
on newer systems. The xwin library is deprecated for new apps -- with
recommendations to use newer toolkits (I'm not that familiar w/any of
them, just saying...). I've yet to get a clean compile off of any
of your releases -- but haven't spent much time investigating either.
Most *nix products I can run configure and it works, though newer
ones based on Cmake from windows-land have given me more grief (have
gotten some to work -- just not as easily).
> [...]
>
> > I meant having meter's setup being done in the meter's
> > constructor.
>
> In xosview2 I removed all non-initialization code from all meters
> constructors. This is because when meters are being constructed the
> window is not even created yet. First all the meters are created,
> then the window is (and sized to fit them). So, the meters don't do
> anything until they are told to configure themselves and then collect
> data. This of course could be changed. But is more of a major
> change to what currently exists.
>
----
Yeah... I may never get my current code to work, but
currently I have the win-system initialize up to the point of
windows being made -- then metermaker calls the meter code to do
that last bit. But not sure how it will look a working version.
I'm trying to get the meters to do all of their init in
their constructor -- then add themselves to "class_registry" that
metermaker can loop through. The class_registry is a meter_registry
for most meters -- but for some meters, they are really a "class" of
meters.. like 'net', since out of 8 network interfaces shown by
'ip link' (inc. lo & a bridge), I only care about the traffic on
2 of them, and not a summary at all, so my display from 2+
years ago, display 'eth2' (my external intf.) and br0, (an
internal-only net).
Something that I did on my meters that I'm not *entirely*
satisfied with is "fairly" free-ranging for some of the meters. Since
right now: br0 is showing a top end of 71K and most measurements <10K,
but it can have traffic up to 600M or so. I like being
able to see the patterns on both, but I need a way for history
to show the meter-range transitions. Ideally, I'd also like
to be able to place cursor over a spot on the meter and it
be able to tell me in a pop-up window the actual speed &
top processes contributing to the stat at that point.
Do you have a PC? Have you tried 'processhacker' (think it
is on sourceforge, ask goog)? Or there's one on linux that's
no longer maintained.(sigh)..."qps" that pops up info windows
when you move the mouse over things.
Oh... forgot.. one thing that I need to take into account
when MM makes meters, is that different 'meters' need to be able to
return different numbers of actual graphs that they want to
put out (like my net case, putting out 2 meters, for 8 intfs).
That's why I am trying to get the meters to self-init at the
start so they can already know how many meters they'll want to
display...
Sorry for the essay... but your's, thankfully, aren't that short
either! ;-)
linda
|
|
From: <mik...@co...> - 2017-05-21 01:44:02
|
>>>>> " " == L A Walsh <xo...@tl...> writes:
> The flexibility isn't really at issue. In fact, to
> maintain compatibility, any initial splitting off of the config
> would likely start by duplicating parts of the Xrm.
If X resources were replaced, then the sky is the limit. No reason
why any existing interface would need to be kept. All that is needed
is some clean way for users to specify options and the code to use
them. The key thing is that the interface be clean and easy to use
for both. :)
[...]
> But however & wherever the configs are,
> currently, they are mixed in with display parameters specific
> to X. How they are mixed in and the flexibility of that system
> doesn't seem to be an essential feature of an OS-monitor
> program.
Not sure I'm following you exactly here. Are you talking about the
distinction between what to display and how to display it? As I see
things the configurable options can be broken down into two basic
categories:
1) What information to display: (cpu, memory, swap, etc).
2) How to display it: (color, type of meter, maximum values, fonts,
etc)
I suppose #2 could be broken down even further into pure "graphics"
type options (color, font, etc) and everything else. Regardless of
how one would subdivide things, I'm not sure of the benefit unless the
data collecting code were split from the display code in some way.
Which is what I think you are driving at.
Doing such a split would probably best be done by first having a
second viable display system to use first. Then one could see if
there was some good way to get information from a data collecting part
of the code to the display part of the code. Doing such a split
while there is only one display system would probably result in a
split that would have to be done again when a second display system
was actually added. I know you have suggested some sort of web
display system. My thoughts on that are below.
[...]
> I'm not a web expert, but it seems that one might have a
> suitable graphics framework using an HTML5-based canvas (?).
> If that were the case, If XOSV had meter and display functions
> as separate things, it might be more eaily extensible to those
> areas. Again,
I have thought about this from time to time. As I see it,
displaying system related information as a "web app" would require a
few things:
1) Some code which runs on the actual hardware being monitored.
Because of security concerns this code would not be loaded from a
http server. Probably it would have to be some sort of "stats"
daemon running on the hardware already.
2) Some network interface for this daemon/server to talk to clients
with. Since the client will be a web browser this interface
would probably best be an http server. Which would mean the code
in #1 would either need to implement an http server or be part of
an existing one.
2) Some way to load "client code" into the web browser (html,
javascript, etc). I'm not that familiar with html5 canvases.
But I think this would involve html and probably javascript.
3) Some way to get the information from #1 into the #2. There are
some http/html methods I'm aware of to do this (SSE, etc). So,
it is probably possible to open a socket that stays open for a
while between a server and an http client. But how long it stays
open may depend a lot on both the server and client browser. It
would probably also be best to have this connection use UDP
packets. But I'm not aware of any web standards in that regard.
One could probably do some sort of proof of concept of the above by
using something like cherrypy to server up a simple http server which
would send out cpu stats to a html page with an html5 canvas. I'm
not sure how good or bad the performance of something like this would
be. If the network connection is reasonable, X itself works ok
remotely. So, in theory it should be possible. Just not sure how
much of an unknown a web browser would be.
> FWIW, I have no immediate plans to try to port
> it elsewhere, just thinking about modular & OO design (maybe
> from reading Stroustrup's 2011 version of C++).
It will be nice once c++11 compilers are the default on most
platforms. Last I checked (about a year ago), c++98 compilers were
still the default on many platforms. This may change in the next
year or two. Until then, I am hesitant to use any c++11 or newer
features since it would require many folks to first install a compiler.
[...]
> I meant having meter's setup being done in the meter's
> constructor.
In xosview2 I removed all non-initialization code from all meters
constructors. This is because when meters are being constructed the
window is not even created yet. First all the meters are created,
then the window is (and sized to fit them). So, the meters don't do
anything until they are told to configure themselves and then collect
data. This of course could be changed. But is more of a major
change to what currently exists.
Mike
|
|
From: L A W. <xo...@tl...> - 2017-04-21 17:26:09
|
mik...@co... wrote:
>> L A Walsh <xo...@tl...> writes:
>>
>
> > I was wondering if anyone has given any thought to splitting
> > xosview's meter config off into a .xosviewrc?
>
> The X resource machinery is pretty flexible.
----
The flexibility isn't really at issue. In fact, to maintain
compatibility, any initial splitting off of the config would likely
start by duplicating parts of the Xrm.
> There are about a dozen or so ways to specify X resources for an
> application. And xosview supports most if not all of them.
>
----
Yes, I'm aware of them. Right there, IMO, is something that
physically, might best reside in the XOSV project, but conceptually,
would likely be best split away from "meter code". I see the 5-15 ways
to store X-prefs, as an "X-thing", that should likely have been handled
internally to Xrm -- w/at most, 'client' code passing in the various
paths of DB's it might want to merge. But however & wherever the
configs are, currently, they are mixed in with display parameters
specific to X. How they are mixed in and the flexibility of that
system doesn't seem to be an essential feature of an OS-monitor
program.
FWIW, I use X all the time, as my main *nix platform, wasn't designed
for graphics, but was supplied with a onboard video subsystem to support
a text-based VGA-based console. So I do most of my work on that machine
remotely, usually using an X-based gvim for editing.
> > maybe another way of putting that is has anyone given any
> > thought to how config data would be structured if the display
> > system was to change... say, using an html canvas or some other
> > non-X based display.
>
> If xosview were to not link with the X libraries then some
> replacement for dealing with configuration would need to be either
> used or written from scratch. But, of course xosview could render
> using something like opengl (I've been experimenting with this) but
> still use the X resource mechanism for parsing configuration.
>
> The nice thing about X is that it is available over a very wide
> range of operating systems. Other rending and windowing systems are
> much more platform specific.
Except, as an example, the web -- which is available on even more
systems
than X. While it might be possible to port X to an android based OS, for
example, I'm not sure how well it would perform or be useful (FYI, I don't
have a smartphone, so not trying to promote a specific platform). Other
small players, at this point, would be display systems like Wayland (in
which
they are striving to not be dependent on X) and "rendering" systems like
'NX' used mostly in transport & rendering of remote desktops. There
may be others, but those are the only one's I've thought of.
I'm not a web expert, but it seems that one might have a suitable
graphics framework using an HTML5-based canvas (?). If that were the case,
If XOSV had meter and display functions as separate things, it might be
more eaily extensible to those areas. Again, FWIW, I have no immediate
plans to try to port it elsewhere, just thinking about modular & OO
design (maybe from reading Stroustrup's 2011 version of C++).
>
> I assume the same thing would apply to
> various config file parsing libraries. So, a replacement for X
> resources would probably be most portable if written from scratch and
> included with the xosview source.
>
----
I'd think so. At least that's what I was thinking at this
point (subject to change if confronted w/other evidence).
> > I'm wondering how good it is to bundle the meter config into the
> > X display system.
>
> As I alluded to above, I don't think it is much of a negative
> point. Particularly since xosview does not support any other
> windowing system other than X. If at some point it does then the
> need and or benefit of using a different configuration library would
> be more applicable.
>
Better to be thinking and or planning things in advance so
if/when need knocks at your door, you've at least got some ideas...
> Call me Mr. Skeptical, but I don't see another portable windowing
> system other than X becoming wide spread anytime in the near future.
> I did attempt to research wayland a while back. Not only is it's use
> not wide spread, there does not seem to be much documentation or
> client code examples to work with.
>
----
The widest spread GUI interface today is the web, but if someone
wants to remotely monitor their system it wouldn't just be about display
but also transport. For that matter, one might want to transport
data and not graphics. You'd also want something that could easily
restart/reconnect after being asleep. X doesn't do that so well in my
experience. I don't know if TCP would be that good for a start/stop
type of usage, or UDP, but not really something I've thought about much.
> I have done some work with xosview2 to split up the data gathering
> code from the rendering code. And there is a possibility that I might
> add optional support for xosview to render using opengl. This may cut
> down on the resource usage (of the cpus) by xosview. Work would need
> to be done re-factoring the window specific parts of xosview first.
> And the opengl support would be an optional compile time configuration.
>
resource usage of the cpu's? Don't usually see much cpu usage.
Possible reason -- My X display is on a different machine than the
machine being measured. So for cpu-usage, I only saw the usage needed
to read the measurements. Anything related to rendering cost
happens on a different machine. I focused more on bandwidth issues,
primarily because I was interested in network usage on 2 of
my ethernet interfaces (one for internal traffic, other for external).
Also focused a bit on aesthetics -- like filling each meter.
It bugged me having it start by only using a little bit of
the meter on the right and taking "a while" to fill in the
rest of the meter. So I start by filling in the entire width of
the meter with the first sample.
The 2nd would take half, 3+4 would take a fourth.. until the meter
was filled w/measurements when it starts doing simple left shifts,
a pixel at a time. The initial filling used the same
shifting of an area used later to shift things over a pixel
at a time.
> > Along those lines, it seems that, at least in C++, it might be
> > convenient for a per-meter config routine to setup all of the
> > wanted meters before main set's up the X-display, but to do that
> > would require the meters storing their config elsewhere.
>
> Pretty much every meter can and is configured before the X window
> pops up.
---
I meant having meter's setup being done in the meter's constructor.
> There is one aspect of this configuration that is still on
> my TODO list. It would be nice to allow the order meters are arranged
> to be user configurable.
---
Yeah... only thing I did along those lines were in the CPU
area where the user could specify order and combining, with default
being to separate by mandatory and discretionary usage -- mandatory --
usage caused by interrupts or cycle-stealing, with discretionary
being the more standard USR+SYS+NICE. Have the discretionary at
the bottom, IDLE in middle, and mandatory @ top.
> Right now a meter can be turned on or off
> via an X resource. But the order the meters appear in the window is
> hard coded in MeterMaker. Allowing this order to be customized would
> not be too much work. I've just not had time to do it.
>
>
> > Another Q, is have you been doing any annotations for doxygen,
> > or did I see something about it getting removed? Was it
> > replaced with some other system?
>
> Personally, I have not plans to add any doxygen related markup to
> xosview2.
>
----
Ahh, so you've never used it before either**. I hadn't before
this project, but was looking for something to help me see how
things interrelated. It shows that with *no* markup -- just
by code analysis. The only thing I did for extra info was
inserting a comment before a function where the 1st line starts
with '///', w/subsequent lines prefixed by '//'. A blank line
ends the comment. Those types of comments are included with
the function as its description.
I think there are a bunch of other things that _could_ be
done, but I was interested more in the automatic stuff, and with
my adding comments before a func (in a minority of cases where
things might not be clear).
** -- I say that because I had no interest in putting foreign
looking formatting that might tend to look like a wart on what
could potentially be elegant looking C++ code, which sounds
similar to your take. I didn't realize how much the tool
did by itself w/nothing more than a config file.
Attached the graph of xos2.1 from last fall...
But in addition to the graphs, it can show source code,
symbol cross refs -- all hyperlinked to make it easy to
follow a chain -- anyway, I thought it was a neat new
toy, er, software development tool... :-)
Anway, if I get anything working, I'll send it by.
Haven't worked on it since you took things over again, so
the code is a bit "cold" (did I write that?)... oh well..
-linda
|