RFR [XS]: 8244196: adjust output in os_linux

[Stefan Karlsson]

On 2020-06-09 07:21, David Holmes wrote:
> Hi Stefan,
>
> On 9/06/2020 4:21 am, Stefan Karlsson wrote:
>> Hi all,
>>
>> I encountered the output of this today, and I think the current 
>> output looks pretty bad. From my hs_err file I have this:
>>
>> CPU:total 32 (initial active 32) (8 cores per cpu, 2 threads per 
>> core) ... much more ...
>> Online cpus:
>> 0-31
>> Offline cpus:
>>
>> cache level:
>> 1
>> cache type:
>> Data
>> cache size:
>> 32K
>> cache coherency line size:
>> 64
>> cache level:
>> 1
>> cache type:
>> Instruction
>> cache size:
>> 32K
>> cache coherency line size:
>> 64
>> cache level:
>> 2
>> cache type:
>> Unified
>> cache size:
>> 256K
>> cache coherency line size:
>> 64
>> cache level:
>> 3
>> cache type:
>> Unified
>> cache size:
>> 20480K
>> cache coherency line size:
>> 64
>> BIOS frequency limitation:
>> <Not Available>
>> Frequency switch latency (ns):
>> 0
>> Available cpu frequencies:
>> <Not Available>
>> Maximum cpu frequency:
>> 3200000
>> Minimum cpu frequency:
>> 1200000
>> Current cpu frequency:
>> 2799113
>> Available governors:
>> performance powersave
>> Current governor:
>> powersave
>> Core performance/turbo boost:
>> <Not Available>
>>
>> It's not easy at all to read. Would you mind if I changed this to:
>> Online cpus: 0-31
>> Offline cpus:
>> cache level: 1
>> cache type: Data
>> cache size: 32K
>> cache coherency line size: 64
>> cache level: 1
>> cache type: Instruction
>> cache size: 32K
>> cache coherency line size: 64
>> cache level: 2
>> cache type: Unified
>> cache size: 256K
>> cache coherency line size: 64
>> cache level: 3
>> cache type: Unified
>> cache size: 20480K
>> cache coherency line size: 64
>> BIOS frequency limitation: <Not Available>
>> Frequency switch latency (ns): 0
>> Available cpu frequencies: <Not Available>
>> Maximum cpu frequency: 3200000
>> Minimum cpu frequency: 1200000
>> Current cpu frequency: 1870913
>> Available governors: performance powersave
>> Current governor: powersave
>> Core performance/turbo boost: <Not Available>
>>
>> I don't think we should be too afraid of changing the lines here. 
>> We've have been changing other parts of the hs_err files, and if 
>> anyone is scraping this info, I think they'll need to adapt.
>
> This is mostly my fault. I was focusing on the changes to 
> os::Linux::print_proc_sys_info which originally removed a lot of 
> newlines, which I thought may cause issues for anyone scraping the 
> info. So Matthias fixed that. But the fix moved the newline printing 
> inside _print_ascii_file_h which then caused all the extra newlines 
> that you are complaining about. :( Sorry about that.
>
> Given I was concerned about simply removing the newline in all cases 
> before, I have to remain concerned about removing them all now. I 
> think we need to restore this:
>
>   static void _print_ascii_file_h(const char* header, const char* 
> filename, outputStream* st) {
> !   st->print("%s", header);
>
> (rather than st->print_cr) and then add the \n in the callers that 
> want the header on a line by itself.

OK. Thanks! I'll prepare a patch to clean this up for JDK 15.

StefanK

>
> David
> -----
>
>> Thanks,
>> StefanK
>>
>> On 2020-04-30 17:46, Baesken, Matthias wrote:
>>> Hello, please review this small change to os_linux.cpp  .
>>>
>>> It  uses the  recently introduced    _print_ascii_file_h
>>> Function t more places (makes the code  a bit  shorter ) .
>>>
>>>
>>> Bug/webrev :
>>>
>>> https://bugs.openjdk.java.net/browse/JDK-8244196
>>>
>>> http://cr.openjdk.java.net/~mbaesken/webrevs/8244196.0/
>>>
>>> Thanks, Matthias
>>