[rteval] Fails to work without numactl

[rteval] Fails to work without numactl

[Sebastian Andrzej Siewior]

rteval as of v3.2 (and the main branch) fails to work continue without
numactl around:

| [DEBUG] [kcompile] systopology: 1 node system (8 cores per node)
| [DEBUG] [kcompile] Configuring build job for node 0
| Exception in thread kcompile:
| Traceback (most recent call last):
|   File "/usr/lib/python3.9/threading.py", line 954, in _bootstrap_inner
|     self.run()
|   File "/home/bigeasy/rteval/rteval/modules/__init__.py", line 181, in run
|     self._WorkloadSetup()
|   File "/home/bigeasy/rteval/rteval/modules/loads/kcompile.py", line 206, in _WorkloadSetup
|     self.buildjobs[n] = KBuildJob(self.topology[n], self.mydir, \
|   File "/home/bigeasy/rteval/rteval/modules/loads/kcompile.py", line 55, in __init__
|     self.binder = 'taskset -c %s' % compress_cpulist(cpulist)
|   File "/home/bigeasy/rteval/rteval/misc.py", line 62, in compress_cpulist
|     if isinstance(cpulist[0], int):
| TypeError: 'NoneType' object is not subscriptable

It continues to wait for kcompile…

With numactl around it uses a different path and continues. However it
then finishes (the main binary) while the compile job and hackbench
continues (not sure if this is intended).

Sebastian

Re: [rteval] Fails to work without numactl

[Sebastian Andrzej Siewior]

On 2021-08-16 13:58:09 [+0200], To linux-rt-users@vger.kernel.org wrote:
> rteval as of v3.2 (and the main branch) fails to work continue without
> numactl around:
…
> | TypeError: 'NoneType' object is not subscriptable
> 
> It continues to wait for kcompile…
> 
> With numactl around it uses a different path and continues. However it
> then finishes (the main binary) while the compile job and hackbench
> continues (not sure if this is intended).

ping.

Sebastian

Re: [rteval] Fails to work without numactl

[John Kacur]

[-- Attachment #1: Type: text/plain, Size: 1415 bytes --]



On Mon, 30 Aug 2021, Sebastian Andrzej Siewior wrote:

> On 2021-08-16 13:58:09 [+0200], To linux-rt-users@vger.kernel.org wrote:
> > rteval as of v3.2 (and the main branch) fails to work continue without
> > numactl around:
> …
> > | TypeError: 'NoneType' object is not subscriptable
> > 
> > It continues to wait for kcompile…
> > 
> > With numactl around it uses a different path and continues. However it
> > then finishes (the main binary) while the compile job and hackbench
> > continues (not sure if this is intended).
> 
> ping.
> 
> Sebastian
> 

What distro are you running on Sebastian?
In rpm based distros numactl is a Requires, so that path where it is not 
available is rarely tested - but that doesn't mean I won't take a patch or 
try to fix it myself. I did have a quick look at it.

kcompile.py compiles an allmodconfig kernel. (not sure this is the best 
thing to do still, but that's a different topic)

The requirements for that continues to grow. I have often seen rteval fail
when requirements for building the kernel aren't installed on a machine.

You can usually figure that out by looking at the debug logs in rteval, 
but another simple thing to do is simply try an allmodconfig kernel 
compile on your machine and when it fails figure out what 
software is needed. It wouldn't be a bad thing to have a list of Debian 
build requirements and Fedora build requirements, etc.

John

Re: [rteval] Fails to work without numactl

[Sebastian Andrzej Siewior]

On 2021-08-30 16:33:32 [-0400], John Kacur wrote:
> What distro are you running on Sebastian?
Debian but I didn't install it in anyway. I just run make of the git
checkout and this started it if I remember correctly.

> In rpm based distros numactl is a Requires, so that path where it is not 
> available is rarely tested - but that doesn't mean I won't take a patch or 
> try to fix it myself. I did have a quick look at it.

From the checkout there were complains about other libs but nothing
complained about numactl. Thanks.

> kcompile.py compiles an allmodconfig kernel. (not sure this is the best 
> thing to do still, but that's a different topic)
> 
> The requirements for that continues to grow. I have often seen rteval fail
> when requirements for building the kernel aren't installed on a machine.

okay.

> You can usually figure that out by looking at the debug logs in rteval, 
> but another simple thing to do is simply try an allmodconfig kernel 
> compile on your machine and when it fails figure out what 
> software is needed. It wouldn't be a bad thing to have a list of Debian 
> build requirements and Fedora build requirements, etc.

Okay.

> John

Sebastian