abidiff: Added/Deleted/Changed markers

[Dodji Seketeli]

Hello,

Matthias Männich <maennich@google.com> a écrit:

> Hi!
>
> Thanks for looking into this!
>
> On Tue, Mar 03, 2020 at 11:53:56AM +0000, Giuliano Procida wrote:
>>Added/Deleted/Changed markers appear in abidiff output under various
>>circumstances.
>>
>>In --leaf-changes-only mode, "[A]", "[D]", "[C]" appear
>>unconditionally to prefix each of the changes (and notably, this is
>>what we see when doing kernel ABI diffs). In normal mode, they only
>>appear if there are more than 100 changes in the given section.
>>
>>"[C]" is also not always followed by a space character.
>>
>>The markers are redundant in that they can be determined from their
>>context. They would be most useful if doing something like
>>line-by-line processing of abidiff output, but they are not always
>>emitted so that's not happening.
>>
>>I'd like to do one of two things to make abidiff output more
>>consistent and predictable. Either option will require the scripted
>>adjustment of test cases. It's straightforward (I've already done it).
>>
>>1. Emit them unconditionally, always followed by a space character.
>>2. Remove them altogether.
>
> I feel like option 1) would preserve the current behaviour and would
> make it consistent. Dodji, any particular reason for the 100 threshold?
> If not, I would suggest the 'emit unconditionally' way of addressing
> this.

It's true that the markers are redundant in that they can be determined
from their context.  The problem though is that when there are many
changes (many lines) on the screen, the user cannot see the context
without scrolling.

Some users reported this little annoyance to me many years ago, so
that's why we decided to just add the markers when they were "many"
changes.

Now I guess that users who want to grep the output for various reasons,
would want more predictability (consistency) here.

I don't have a strong opinion, so I'll agree with whatever you guys find
more useful. I would tend to go for the "emit unconditionally" as well
and say that redundancy is not a real issue in this case.  But that's
just me.

Cheers,

-- 
		Dodji