Re: [PATCH] CHANGELOG.md: Make PV shim smaller by factoring out HVM-specific shadow code

[George Dunlap <George.Dunlap@xxxxxxxxxx>]

> On Mar 29, 2021, at 6:26 PM, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> wrote:
> 
> On 29/03/2021 17:23, Jan Beulich wrote:
>> On 29.03.2021 18:12, George Dunlap wrote:
>>>> On Mar 25, 2021, at 7:08 AM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>>> On 24.03.2021 18:26, George Dunlap wrote:
>>>>> Signed-off-by: George Dunlap <george.dunlap@xxxxxxxxxx>
>>>>> ---
>>>>> Missed one from my list when creating the other series
>>>>> 
>>>>> CC: Ian Jackson <ian.jackson@xxxxxxxxxx>
>>>>> CC: Jan Beulich <jbeulich@xxxxxxxx>
>>>>> CC: Roger Pau Monne <roger.pau@xxxxxxxxxx>
>>>>> ---
>>>>> CHANGELOG.md | 1 +
>>>>> 1 file changed, 1 insertion(+)
>>>>> 
>>>>> diff --git a/CHANGELOG.md b/CHANGELOG.md
>>>>> index 15a22d6bde..49832ae017 100644
>>>>> --- a/CHANGELOG.md
>>>>> +++ b/CHANGELOG.md
>>>>> @@ -18,6 +18,7 @@ The format is based on [Keep a 
>>>>> Changelog](https://keepachangelog.com/en/1.0.0/)
>>>>> - x86_emulate: Expanded testing for several instruction classes
>>>>> - CI loop: Add Alpine Linux, Ubuntu Focal targets; drop CentOS 6
>>>>> - CI loop: Add dom0less aarch64 smoke test
>>>>> + - Factored out HVM-specific shadow code, allowing PV shim to be slimmer
>>>> But shadow code doesn't get included by default in shim-exclusive
>>>> builds (and others are unlikely to disable HVM).
>>> Can you propose some better text please?
>> Does this need mentioning here in the first place?
> 
> I would recommend not.
> 
> We've been doing incremental improvements for the shim for several
> releases now, and in this case, we're literally talking a few kb of
> code.  As we already align to 2M boundaries for superpage reasons, there
> almost certainly isn't actually a reduction in runtime size.

I don’t understand why the two of you are downplaying your work so much.  Yes, 
these are all only incremental improvements; but they are improvements; and the 
cumulative effect of loads of incremental improvements can be significant.  
Communicating to people just what the nature of all these incremental 
improvements are is important.

I mean, look at the release notes for Go 1.15 [1].  It includes things like 
this:

"JSEscape now consistently uses Unicode escapes (\u00XX), which are compatible 
with JSON."

"go test -v now groups output by test name, rather than printing the test name 
on each line."

Those sound far more trivial than “Even more shadow code has been moved to an 
HVM-specific file”.

If the approach is going to be “SUPER IMPORTANT SPECIAL STUFF ONLY", I’d 
recommend removing CHANGELOG.md.  Having an official list that says, “Well, 
really, we only did 2 things this release” is going to be actively harmful.  

 -George

[1] https://golang.org/doc/go1.15