On Fri, Sep 15, 2023 at 2:31 PM Mario Limonciello mario.limonciello@amd.com wrote:
On 9/15/2023 16:15, Doug Smythies wrote:
On 2023.09.15 03:41 Swapnil Sapkal wrote:
In intel_pstate_tracer.py, Gnuplot is used to generate 2D plots. In current implementation this tracer gives error while importing the module because Gnuplot is imported from package Gnuplot-py which does not support python 3.x. Fix this by using pygnuplot package to import this module.
As described in the prerequisites section, the package name is distribution dependant. On my distribution the original package name is phython3-gnuplot, and it is working fine.
sys.version_info(major=3, minor=8, micro=10, releaselevel='final', serial=0)
I don't currently have python3-pygnuplot installed, and so this patch breaks the intel_pstate_tracer for me.
So, I installed the python3-pygnuplot package, and it still didn't work, as there still wasn't a pygnuplot module to import. So, I found something called PyGnuplot.py and so changed to that and got further. But then it got upset with:
File "./intel_pstate_tracer.py.amd", line 298, in common_gnuplot_settings g_plot = gnuplot.Gnuplot(persist=1) NameError: name 'gnuplot' is not defined
I gave up and returned to the unpatched intel_pstate_tracer.py And checked that is still worked fine. It did.
So, I do not accept this proposed patch.
Not really related, but for a few years now I have been meaning to change the minimum python version prerequisite to >= 3.0 and to change the shebang line from this:
#!/usr/bin/env python
To this:
#!/usr/bin/env python3
I have to use the latter version on my distro. Back when I looked into it, things were inconsistent, so I didn't know what to do. The kernel tree has 52 .py files of the latter shebang and 11 of the former.
... Doug
Presumably this is the one that Swapnil intended:
Yes, I found that earlier. For my part of it, I do not want to use any out-of-distro package.
It requires python3, so I think if upgrading to this one the script does need to be switched to python3. Besides the shebang, you should also use a helper like 2to3 to look for any other changes.
I already did the python 3 patch in January, 2020: commit e749e09db30c38f1a275945814b0109e530a07b0 tools/power/x86/intel_pstate_tracer: changes for python 3 compatibility
I haven't had any issues since, shebang aside.
... Doug
There were 97 hits for 'gnuplot' at pypi. 2 stood out but at least in the case of gnuplot based stuff, I think it's worth dropping a comment that links back to pypi page for the intended package.
Another alternative is to include a 'requirements.txt' file that pip can pick up.
https://pip.pypa.io/en/stable/reference/requirements-file-format/
Signed-off-by: Swapnil Sapkal swapnil.sapkal@amd.com
tools/power/x86/amd_pstate_tracer/amd_pstate_trace.py | 1 - tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py | 4 ++-- 2 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/tools/power/x86/amd_pstate_tracer/amd_pstate_trace.py b/tools/power/x86/amd_pstate_tracer/amd_pstate_trace.py index 2448bb07973f..14f8d81f91de 100755 --- a/tools/power/x86/amd_pstate_tracer/amd_pstate_trace.py +++ b/tools/power/x86/amd_pstate_tracer/amd_pstate_trace.py @@ -27,7 +27,6 @@ import re import signal import sys import getopt -import Gnuplot from numpy import * from decimal import * sys.path.append(os.path.join(os.path.dirname(__file__), '../intel_pstate_tracer')) diff --git a/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py
b/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py
index ec3323100e1a..68412abdd7d4 100755 --- a/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py +++ b/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py @@ -32,7 +32,7 @@ import re import signal import sys import getopt -import Gnuplot +from pygnuplot import gnuplot from numpy import * from decimal import *
@@ -295,7 +295,7 @@ def common_all_gnuplot_settings(output_png): def common_gnuplot_settings(): """ common gnuplot settings. """
- g_plot = Gnuplot.Gnuplot(persist=1)
- g_plot = gnuplot.Gnuplot(persist=1)
# The following line is for rigor only. It seems to be assumed for .csv files g_plot('set datafile separator ","') g_plot('set ytics nomirror') -- 2.34.1