Re: [XEN PATCH 9/9] CI: Add timing to junit

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 04/06/2025 9:48 am, Anthony PERARD wrote:
> On Tue, Jun 03, 2025 at 11:35:22AM -0700, Stefano Stabellini wrote:
>> On Tue, 3 Jun 2025, Anthony PERARD wrote:
>>> From: Anthony PERARD <anthony.perard@xxxxxxxxxx>
>>>
>>> Signed-off-by: Anthony PERARD <anthony.perard@xxxxxxxxxx>
>>> ---
>>>  automation/scripts/run-tools-tests | 5 ++++-
>>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/automation/scripts/run-tools-tests 
>>> b/automation/scripts/run-tools-tests
>>> index 852c1cfbcf..e38cc4068c 100755
>>> --- a/automation/scripts/run-tools-tests
>>> +++ b/automation/scripts/run-tools-tests
>>> @@ -18,9 +18,12 @@ for f in "$1"/*; do
>>>          continue
>>>      fi
>>>      echo "Running $f"
>>> -    printf '  <testcase name="%s">\n' "$f" >> "$xml_out"
>>> +    time_start=$EPOCHREALTIME
>>>      "$f" 2>&1 | tee /tmp/out
>>>      ret=${PIPESTATUS[0]}
>>> +    time_end=$EPOCHREALTIME
>>> +    time_test="$(bc <<<"$time_end - $time_start")"
>> As it looks like no other scripts need bc at the moment but we already
>> rely on awk (automation/scripts/xilinx-smoke-dom0less-arm64.sh) I'd
>> prefer this version:
>>
>> time_test="$(awk "BEGIN {print $time_end - $time_start}")"
> You mean I have to choose between busybox and busybox?
>
> $ ls -l $(which bc awk)
> lrwxrwxrwx    1 root     root            12 Feb 13 23:19 /usr/bin/awk -> 
> /bin/busybox
> lrwxrwxrwx    1 root     root            12 Feb 13 23:19 /usr/bin/bc -> 
> /bin/busybox
>
> :-)
>
> I guess it doesn't really matter.
>
> One difference though:
> $ awk "BEGIN {print $time_end - $time_start}"
> 3.28798
> $ bc <<<"$time_end - $time_start"
> 3.287982
>
> awk is less precise, but I guess that doesn't matter as well, gitlab UI
> isn't going to show the extra digits.
>
> So I guess I can change to use `awk` instead, just in case for some
> reason `bc` isn't present, since `awk` seems mandatory for many of our
> scripts.
>
> Thanks,

bc is a normal posix utility just like awk is, so there's no change in
dependencies caused by this.

Furthermore, it is the right tool for this job in a way that awk isn't. 
Besides the reduction in precision, see xen/tools/check-endbr.sh for the
number of ways we found that awk malfunctions with large numbers.

I firmly suggest continuing to use bc here.

~Andrew