Hi all,

Here is my bisect dunno patch series again.
The changes since last time are the following:

[PATCH 1/7] rev-list: implement --bisect-all
[PATCH 2/7] Bisect: fix some white spaces and empty lines breakages.

-> No change.

[PATCH 3/7] Bisect: implement "bisect dunno" to mark untestable revisions.

-> Added dunno stuff in "bisect_replay" that I had forgotten.
-> Use "bisect_write_good" and "bisect_write_bad" in "bisect_replay" 
while at it.

[PATCH 4/7] Bisect: factorise "bisect_write_*" functions.
[PATCH 5/7] Bisect: factorise some logging into "bisect_write".
[PATCH 6/7] Bisect: factorise "bisect_{bad,good,dunno}" into "bisect_state".

-> Some new factorisation and clean up work.

[PATCH 7/7] Bisect: add "bisect dunno" to the documentation.

-> Document "bisect dunno" and fix some short usage descriptions.

Regards,
Christian.

-

[ message continues ]

Good work on the series, Christian, but don't you think that  
"unknown" would sound a little bit better than "dunno"? For people  
who don't speak English as a second language "dunno" might not be  
immediately clear.

Cheers,
Wincent

-

[ message continues ]

Doh, I meant to say "people who don't speak English as a *first*  
language".

Cheers,
Wincent


-

[ message continues ]

"undecided"?

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

I choosed "dunno" because that was what Dscho suggested in this thread:

http://thread.gmane.org/gmane.comp.version-control.git/53584/focus=53595

It seems to me short and understandable at the same time.

More meaningfull would be "untestable" or "cannottest" or "canttest" but 
it's much longer, while "good" and "bad" are short.

Christian.


-

[ message continues ]

I would think that tongue-in-cheek.  In case it was serious, I'd
consider it one of those cases where it would make good sense to
overrule the geek penchant for quirkiness.

Just think about the nuisance of finding adequate translations in
i18n: "Woaßnet", "Eh?", "Chepas" etc.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

Well, that won't be a problem, as commands and their subcommands and
options aren't translated.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231
-

[ message continues ]

Ugly?  Neutral?

René
-

[ message continues ]

"Ugly" has a certain charm to it but would probably not translate
well.  "Limbo" would be another such candidate, probably with better
translatability.  But while some of those have some geeky appeal, I
really think something reasonably plain like "undecided" would be
better in the long run.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

"undecided" sounds good to me. It should be clear to non-native  
speakers of English (at least, clearer than "dunno").

<personal opinion>
   Another problem with "dunno" is that it might come across as a bit  
unprofessional due to its informality. Yes, you'll find it in most  
dictionaries, but always with a qualifier of "slang", "colloquial",  
"casual", "pronunciation spelling" or similar.
</personal opinion>

Cheers,
Wincent

-

[ message continues ]

What about just "unknown"?

--=20
=2Emarius

[ message continues ]

I tend to nitpick to the degree of silliness when my own suggestions
are concerned, but "unknown" sounds to me like the state _before_ the
test.  If a person says he is "undecided" about something that means
that he _has_ thought about it already.  "Undecidable" might bring
this distinction across more strongly, but it is a more complicated
word and it insinuates that it is _impossible_ to come to a decision
regardless of the spent effort.

"unknown" clearly is much better than "dunno" though even if my own
favorite would be "undecided".

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

What then about a good'ol programming favorite, "void"? :-)

I agree that "unknown" might be a state even _before_ a person has=20
determined if a case is good or bad (same for 'dunno' actually: "- Do=20
you know if it works? - I dunno yet") When I think more about it, I=20
really like "void"..

"Argh, this test is void, because someone messed with it"
"We can't make heads or tails of this one, so it must be void"

--=20
=2Emarius

[ message continues ]

"skip"? That would make semantic sense, right?


Dave.
-

[ message continues ]

...or we could go all spaghetti western, and call it "ugly".

(as in "git-bisect [the <good>, the <bad> and the <ugly>]")


Have fun! :)

...Johan

-- 
Johan Herland, <johan@herland.net>
www.herland.net
-

[ message continues ]

<personal opinion>
   Yes, it's funny, but I don't think an SCM interface is a place for  
jokes or puns. Git already has one big tongue-in-cheek attribute:  
it's name, so let's leave it at that.
</personal opinion>

Cheers,
Wincent

-

[ message continues ]

That's also why I suggested "skip"; you might not be able to test a
particular commit, but you might also not *want* to test a particular
commit for some reason.


Dave.
-

[ message continues ]

Skip seems a great choice: it directly expresses the wish to
not consider a certain commit. The reason is unimportant.

-

[ message continues ]

But it is an _action_, while "good" and "bad" are properties.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

"skipped", then. Either way, something like this has got to be much
better than "dunno".


Dave.
-

[ message continues ]

"good" and "bad" are descriptive.  "to be skipped" would be necessary

"undecided" still has my vote, and I could live with "unknown".
Everything that has been proposed since then is, in my opinion,
strictly worse.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
-

[ message continues ]

Yet another very short word: void.

I'm thinking about ticket copies that sometimes are marked "void" so you cannot use it.

-- robin

-

[ message continues ]

My vote is for "supercalifragilisticexpialidocious". It's clearly
superior to the 1500 other suggestions in this thread.

-- 
Karl Hasselström, kha@treskal.com
      www.treskal.com/kalle
-

[ message continues ]

(Not intended as an attack on this particular suggestion, by the way.
Sorry if it sounded a bit harsh.)

-- 
Karl Hasselström, kha@treskal.com
      www.treskal.com/kalle
-

[ message continues ]

Well, this has been debated to death, but I actually think that "skip" is 
a good choice, exactly because it's an action.

"good" and "bad" do indeed describe properties of the commit, and are used 
to describe the state of the tree in question. In contrast, "git bisect 
skip" says not somethign about the state of that tree - it talks about 
what we should *do* with that tree.

IOW, I think "git bisect skip" in some sense has more to do with an action 
like "git bisect start", than with "good" or "bad". 

(Yes, "good" and "bad" have an action associated with them too - namely to 
start the next bisection event - but they are not named according to the 
action they cause, but because they describe the tree state)

			Linus
-

[ message continues ]

Hi,


Could we, _please_, first decide if the implementation has merits, and 
just apply it as is in that case?  We can rename it whatever anybody likes 
later, and we can paint the bikeshed brown if you want to.

Ciao,
Dscho
-

[ message continues ]

I thought everybody really agreed that being able to skip commits that you 
cannot say good/bad about is a feature worth doing?

Right now we actually have some docs in the man-page about doing that 
avoidance manually, so it's not like it's debatable whether this issue 
comes up. It most definitely does come up.

Does anybody really think it's not a good feature? And I've not seen 
negative comments about the implementation either apart from some small 
details that I think got fixed up already (but maybe the complaints were 
all hidden by the shed color discussions ;)

			Linus
-

[ message continues ]

I figured with something like this, it'd be a lot easier to get the
colour right first, since command UI is harder to repaint if it gets
widely adopted. Anyway, I think the patch itself is a very good
feature.


Dave.
-

[ message continues ]

Yeah, or "avoid".

Christian.
-

[ message continues ]

Hi,


I guess this discussion means that nobody has anything technical to argue 
about.  IOW your patch series is good...

;-)

Ciao,
Dscho

-

[ message continues ]

Hi,


Thanks for doing this.  I think that especially the factorisation adds 
tremendously to the readibility.

Ciao,
Dscho

-

[ message continues ]

I now have this series queued in my pu branch.  It passes the tests
it comes with, and doesn't appear to break anything, but apparently
there is also still some debate about what a dunno should be called
("unknown", "void", "ugly", "dunno", "skip" ...).

-- 
Shawn.
-

[ message continues ]

Hi,


AFAICT these are all bikeshed painting arguments, not technical arguments.  
I was initially opposed to having --bisect-all, wanting to have 
--bisect-dunno <ref>...

But in the end, the people doing the work decide, and therefore I am fine 
with --bisect-all, especially since it seems clean enough for me.

As for all those "dunno is no English"...  I'd first merge the technical 
part (i.e. what you have now in pu), and then let the discussion about 
which synonyms to choose continue, until a consensus is formed about other 
names (if there is a consensus at all!).

IMHO there is no reason to hold of the fine work of Christian, just 
because there are non-technical arguments still in the air.

I want bisect dunno.  Even if there is another name later.

Ciao,
Dscho

-

[ message continues ]

Hi,


It's also the fine work of Junio as he wrote most of 

Thanks for your kind words,
Christian.



-

[ message continues ]