[noob] is this normal behavior

[noob] is this normal behavior

[Harry Putnam]

Setup: Running openindiana (hipster) a solaris 11 open source branch
       git 2.12.2
       
I've recently moved from mercurial to git mainly because git appears
to receive the most development and the heaviest use.

I'm using git with hold over behavior developed from my time using cvs
long ago.

I keep a hierarchy of directories and files that mirrors the structure
on the OS where git is installed [just where ever tracked files are
located on OS] not the entire structure.

I keep a git repo on the same OS that is created by rsyncing  the
BUFFERED hierarchy into a git repo

Many of the BUFFERED files are symlinked to there places in the OS

Some are not because certain files being symlinked seemed to cause
permissions type problems on the OS.  In those cases they are
periodically copied to the BUFFER hierarchy 

So, not to get too tangled up in explaining my setup... tracked files
change in the buffer area or new files are created there.

Periodically I rsync the files in BUFFER over the files in git repo

And record the changes thru normal git operations

git status
git add (when needed)
git commit

,----
| Aside:
| While I would welcome comments or suggestions about using a setup like
| this, that is not the subject of this post.
`----

This is my most recent attempt at learning and using git but I have
done a few tries before.

I'm noticing behavior I do not recall seeing in earlier attempts.

rsyncing the buffer over the git repo is now producing a result where
`git status' shows any changed or added files as modified or new but
all need to be added before a commit can take place, not just the new
ones.

Shouldn't files that changed but are already being tracked just show
up as modified and not need adding?

That is, if I have the file ~/.bashrc, which is really a symlink to
BUFFER/OS/export/home/reader/.bashrc

  And is tracked in git repo as REPO/OS/export/home/reader/.bashrc

When;
  BUFFER/OS/export/home/reader/.bashrc

is edited.

Then  rsynced like below:
  BUFFER <root of rsync is here> /OS/export/home/reader/.bashrc
over
  GITREPO <root of rsync> /OS/export/home/reader.bashrc

Thereby changing the repo file in place

Since that file is already being tracked; shouldn't `git status' show
that file as modified but ready to be committed instead of a file that
is modified but needs to be added before a commit can happen?

Another side of this is that a `git diff FILE' only works before an
`git add .' operation is done.

That is, if I run `git diff FILE' AFTER `git add' .. no diff is
reported, even though it is not committed yet.

So, for example: if I'm committing and in the vi buffer of the commit
and want to see a diff of FILE to aid my log notes.

 git diff FILE will report nothing whatever.

Is that expected behavior?

Re: [noob] is this normal behavior

[Igor Djordjevic]

Hi Harry,

Both behaviours you report are normal, specifically:

On 09/05/2017 15:02, Harry Putnam wrote:
> Shouldn't files that changed but are already being tracked just show
> up as modified and not need adding?
> ...
> Since that file is already being tracked; shouldn't `git status' show
> that file as modified but ready to be committed instead of a file that
> is modified but needs to be added before a commit can happen?

No, it shouldn`t - even though the file you`ve modified is already 
being tracked, it doesn`t have to mean you want the modification 
inside your very next commit, and Git doesn`t force that upon you.

That is where index (or "staging area") comes in handy, allowing you 
to `git add` only the changes you actually want to commit now, 
leaving the others for later decision.

You don`t even have to add all the modifications inside the single 
file at once, for example by using `git add --patch`[1] you can 
select just some of the them, fine tuning what gets committed and when.

With some precaution/steps needed not to commit broken project states 
by accident, and some discipline not to overdo it, this allows you to 
fully focus on the actual work you do, making logically unrelated 
changes in place as you see fit, on the go, and only later organizing 
them into logically grouped commits, through diligent use of `git add`.

> Another side of this is that a `git diff FILE' only works before an
> `git add .' operation is done.
> 
> That is, if I run `git diff FILE' AFTER `git add' .. no diff is
> reported, even though it is not committed yet.
> 
> So, for example: if I'm committing and in the vi buffer of the commit
> and want to see a diff of FILE to aid my log notes.
> 
>  git diff FILE will report nothing whatever.
> 
> Is that expected behavior?

Yes, that is as expected - in the form you`ve given, `git diff` shows 
the differences between your working tree and index (staging area), 
so only changes you haven`t added yet. Once you `git add` the changes 
from the working to the index, there are no more differences to show, 
so no diff is produced.

If you want to see the added changes, what will be included in the 
commit if you make it, you can use `git diff --cached`, as per 
git-diff[2] documentation (--staged can also be used instead, being a synonym 
for --cached, but maybe easier to remember, relating it to "staging 
area").

That option shows differences between the staging area (index) and 
the specific commit - with none provided, it implies your currently 
checked-out position (referred to as HEAD), usually being your 
latest/previous commit, which is exactly what you`re interested in.


As a side note, if you think you don`t need it, you can skip staging 
area and commit all the modifications/deletions without a need of 
adding them first by using `git commit --all`, as per git-commit[3] 
documentation.

Just pay attention that untracked files are not affected, you still 
need to add them first to tell Git to start tracking them, including 
them in the next commit, but that seems to align nicely with your 
expectations already.

I personally find the staging area to be one of the greatest Git 
possibilities, but I do understand beginners getting confused by it, 
as admittedly I once was myself.


In the end, you may want to ask questions like this on Git users 
mailing list[4] on Google Groups, being a a nice place for beginners 
to get answers to their concerns.

[1] https://git-scm.com/docs/git-add
[2] https://git-scm.com/docs/git-diff
[3] https://git-scm.com/docs/git-commit
[4] https://groups.google.com/forum/?fromgroups#!forum/git-users

Regards,
Buga

Re: [noob] is this normal behavior

[Harry Putnam]

Igor Djordjevic <igor.d.djordjevic@gmail.com> writes:

[...] snipped detailed reply
 [To Anyone who lands here on a search: Please read Igor D's full
 answer in this thread... ]

> Just pay attention that untracked files are not affected, you still 
> need to add them first to tell Git to start tracking them, including 
> them in the next commit, but that seems to align nicely with your 
> expectations already.
>
> I personally find the staging area to be one of the greatest Git 
> possibilities, but I do understand beginners getting confused by it, 
> as admittedly I once was myself.

Thank you sir for such a full and complete answer.  Adds to by
understanding in a big way.