Big thank you to Ralph for tracking that down and for teaching me
about strace.
Also thanks to those who reproduced the problem, which was not after
all as general my message subject had advertised.
I recall noticing years ago that Screen tends to start somewhat slowly
as a result of closing 400
+ grep -m 1 'model name' /proc/cpuinfo | awk -F ': ' '{print
$2}';uname -r; screen -v; for i in {1..5}; do ./test.sh; sleep 1; done
Intel(R) Core(TM) i5-7500T CPU @ 2.70GHz
4.19.26-1-lts
Screen version 4.06.02 (GNU) 23-Oct-17
real0m0.139s
user0m0.060s
sys 0m0.079s
real0m0.140s
us
Hi again Frederick,
> % time seconds usecs/call callserrors syscall
> -- --- --- - -
> 99.98 11.762280 22524299524285 close
Ah, here we are. https://savannah.gnu.org/bugs/?55618
--
Cheers, Ralph.
Hi Frederick,
> #!/bin/bash
> SCREEN=/usr/bin/screen
> $SCREEN -dmS x
> time $SCREEN -S x -X setenv A B
> $SCREEN -S x -X quit
Perhaps it's closing more file descriptors now than before?
$ strace -c screen -dmS x
Directory '/run/screens' must have mode 777.
% time
$ bash sc.sh
real0m0,140s
user0m0,044s
sys 0m0,097s
$inxi
CPU: Quad Core Intel Xeon E5-1620 v2 (-MT MCP-) speed/min/max:
1497/1200/3900 MHz Kernel: 5.0.0-arch1-1-ARCH x86_64
Up: 1d 1h 34m Mem: 2791.9/64368.0 MiB (4.3%) Storage: 1.83 TiB (23.8%
used) Procs: 230 Shell: bash 5.0.0 inxi: 3
On 3/8/19 7:15 AM, frede...@ofb.net wrote:
> There is no virtualization on my machines.
>
> Did anyone try to reproduce it using my script?
./bench-test.sh
real 0m0.111s
user 0m0.040s
sys 0m0.070s
Intel Core i7-8550U, Kernel: 5.0.0-arch1-1-ARCH
On 08.03.2019 07:15, frede...@ofb.net wrote:
There is no virtualization on my machines.
Did anyone try to reproduce it using my script?
$ bash archtest.sh
real 0m0,004s
user 0m0,004s
sys 0m0,000s
$ inxi
CPU: Dual Core Intel Core i5-3320M (-MT MCP-) speed/min/max:
1559/1200/3300 MHz
Are you using Arch on bare-metal or in VirtualBox 6.x by chance? I saw similar
slowdowns with VirtualBox 6.0 (and 6.2) when run --headless and similar strace
errors (although with poll/write), e.g.
There is no virtualization on my machines.
Did anyone try to reproduce it using my script?
Thank
On 03/07/2019 03:51 PM, frede...@ofb.net wrote:
> Any ideas what might be causing this? I tried to debug with 'strace -r
> -p' and it looks like most of the time is getting spent in a 'select'
> call.
>
> 0.000171 geteuid() = 1000
> 0.75 getegid() = 1000
Dear Arch,
I've noticed a 20x slowdown on a simple IPC task on a couple of
recently updated Arch systems. Here is a script to exhibit the
problem:
#!/bin/bash
SCREEN=/usr/bin/screen
$SCREEN -dmS x
time $SCREEN -S x -X setenv A B
$SCREEN -S x -X quit
On my laptop, which is a slowe
10 matches
Mail list logo
