git-write-tree strangeness

git-write-tree strangeness

[Steven Walter]

I'm importing a large repository from svn into git with a custom tool,
and I noticed a strange issue with one of the commits.

Upon investigating further, I found that when I ran "git-write-tree"
followed immediately by "git-diff-index <tree>" I had differences.  Does
that mean that git-write-tree incorrectly recorded the index, or do I
misunderstand things?

I just wanted to verify that I have at least some clue what is going on
before I dig too much deeper into this issue.  I'm running
git-1.5.3-rc4; I'll see if it occurs with 1.5.2 and bisect if not.
-- 
-Steven Walter <stevenrwalter@gmail.com>
"A human being should be able to change a diaper, plan an invasion,
butcher a hog, conn a ship, design a building, write a sonnet, balance
accounts, build a wall, set a bone, comfort the dying, take orders,
give orders, cooperate, act alone, solve equations, analyze a new
problem, pitch manure, program a computer, cook a tasty meal, fight
efficiently, die gallantly. Specialization is for insects."
   -Robert Heinlein

Re: git-write-tree strangeness

[Alex Riesen]

On 8/8/07, Steven Walter <stevenrwalter@gmail.com> wrote:
> I'm importing a large repository from svn into git with a custom tool,
> and I noticed a strange issue with one of the commits.
>
> Upon investigating further, I found that when I ran "git-write-tree"
> followed immediately by "git-diff-index <tree>" I had differences.  Does
> that mean that git-write-tree incorrectly recorded the index, or do I
> misunderstand things?

What kind of differences? Different sha1 of content?
I can't simply reproduce it.

Re: git-write-tree strangeness

[Junio C Hamano]

"Alex Riesen" <raa.lkml@gmail.com> writes:

> On 8/8/07, Steven Walter <stevenrwalter@gmail.com> wrote:
>> I'm importing a large repository from svn into git with a custom tool,
>> and I noticed a strange issue with one of the commits.
>>
>> Upon investigating further, I found that when I ran "git-write-tree"
>> followed immediately by "git-diff-index <tree>" I had differences.  Does
>> that mean that git-write-tree incorrectly recorded the index, or do I
>> misunderstand things?
>
> What kind of differences? Different sha1 of content?
> I can't simply reproduce it.

Isn't this simply just a dirty worktree and missing --cached
option?

Re: git-write-tree strangeness

[Steven Walter]

On Wed, Aug 08, 2007 at 12:39:03PM -0700, Junio C Hamano wrote:
> "Alex Riesen" <raa.lkml@gmail.com> writes:
> 
> > On 8/8/07, Steven Walter <stevenrwalter@gmail.com> wrote:
> >> I'm importing a large repository from svn into git with a custom tool,
> >> and I noticed a strange issue with one of the commits.
> >>
> >> Upon investigating further, I found that when I ran "git-write-tree"
> >> followed immediately by "git-diff-index <tree>" I had differences.  Does
> >> that mean that git-write-tree incorrectly recorded the index, or do I
> >> misunderstand things?
> >
> > What kind of differences? Different sha1 of content?
> > I can't simply reproduce it.
> 
> Isn't this simply just a dirty worktree and missing --cached
> option?

It doesn't happen everytime; I haven't found a set procedure to cause
the problem.  Your remark, Junio, combined with my later testing, lead
me to believe that this is not a write-tree problem, but a git-add
problem.

The entire procedure I was following was:

    <change working copy by svn>
    git add .
    git add -u
    git write-tree
    git diff-index <tree>

It seems to me that my expectation that diff-index show no changes was
reasonable, but I misunderstood the implication; I expected diff-index
to always compare the given tree to the index, regardless of the
presence or absence of a '--cached' option.

If I rebuilt git from v1.5.2 and repeated the above commands, diff-tree
showed no differences, as I would expect.  Further, if I then re-ran
write-tree/diff-index from v1.5.3-rc4, I still received no differences.
I believe this result exculpates write-tree.

Was there not a thread recently about a missing optimization in git-add?
If this change was included in v1.5.3-rc4, could it be the cause of my
discrepency?  Suggestions for testing this theory, if it is reasonable?
-- 
-Steven Walter <stevenrwalter@gmail.com>
"A human being should be able to change a diaper, plan an invasion,
butcher a hog, conn a ship, design a building, write a sonnet, balance
accounts, build a wall, set a bone, comfort the dying, take orders,
give orders, cooperate, act alone, solve equations, analyze a new
problem, pitch manure, program a computer, cook a tasty meal, fight
efficiently, die gallantly. Specialization is for insects."
   -Robert Heinlein