Re: [GSoC] Git Blog 3

Christian Couder [off-list ref] 于2021年6月7日周一 下午12:45写道：

On Sun, Jun 6, 2021 at 5:21 PM ZheNing Hu [off-list ref] wrote:

quoted

My third week blog finished:
The web version is here:
https://adlternative.github.io/GSOC-Git-Blog-3/

Thanks!

quoted

-----

## Week3: Meticulousness is very important

### What happened this week
- I found a `git cat-file` bug this week, see:

git cat-file --batch=batman --batch-all-objects
batman
fatal: object 00345b5fe0aa8f45e9d1289ceb299f0489ea3fe1 changed type!?

It seems that Git died for a strange reason: the type of
an object changed? Is my Git object damaged? (Just like
a friend of mine, use `find . -type f -print0 | xargs -0 sed

Maybe: s/, use/ used/
or: s/, use/, who used/

quoted

-i "s/\r//g"` remove all the '\r' of all files in a Git repository,

s/remove/to remove/

quoted

this caused most of his Git commands to fail.) So I tested
it under different linux platforms, they all have this same

s/linux/Linux/

quoted

damage.

Maybe: s/damage/breakage/

quoted

After a series of testing and debugging, I found that
`oid_object_info_extended()` did not get the type of
object.

So I submitted the patch for fix this bug to the Git mailing list in
[[PATCH] [GSOC] cat-file: fix --batch report changed-type
bug](https://lore.kernel.org/git/pull.965.git.1622363366722.gitgitgadget@gmail.com/ (local)),
Peff tell us the essential reason for this bug:

Maybe: s/tell/told/

quoted

In `845de33a5b (cat-file: avoid noop calls to
sha1_object_info_extended, 2016-05-18)`, this patches

s/patches/patch/

quoted

skips the call to `oid_object_info_extended()` entirely when
`--batch-all-objects` is in use, and the custom format does
not include any placeholders that require calling it. The correct
solution is to put checking if `object_info` is empty after

s/put checking/check/

quoted

setting `object_info.typep`.

Finally, thanks to the help of Jeff, I summit final patch in

s/summit/submitted a/

quoted

[[PATCH v2 1/2] [GSOC] cat-file: handle trivial --batch format with
--batch-all-objects](https://lore.kernel.org/git/4af3b958dd056e2162fdc5d7f6600bcedde210b8.1622737766.git.gitgitgadget@gmail.com/ (local)).

Talk of experience as a story: Even experienced programmers
make small mistakes, writing any code requires very careful thinking.
- The checkpoints for rejecting `%(raw)` and `--<lang>` are incorrect.
At Junio’s reminder, I changed the checkpoint from
`parse_ref_filter_atom()` to `verify_ref_format()`. My mentor Christian
pointed out some grammatical errors in the cover letter.
- At the suggestion of Junio, I rebased `0efed94 ([GSOC]
ref-filter: add %(raw) atom)` on `1197f1a (ref-filter: introduce
enum atom_type)`, they have a clash, after resolving the conflict,

Maybe: s/have/had/ and s/clash, after/clash. After/

or: s/they have a clash, after/they clashed. After/

quoted

it's better for me to consider the code I implemented before and
the code I wrote now at the same time, I can find more problems
and find better solutions.
- I submitted the patch about `%(rest)`, `%(raw:textconv)` and
`%(raw:filters)` for `ref-filter`, they are used to simulate some

s/, they/. They/

quoted

functions of `git cat-file`, my mentor Hariom noticed one of the

s/, my/. My/

quoted

formatting issues, I am waiting for more reviews for the time being.

### What's the next step

As long as new atoms `%(rest)`, `%(raw:textconv)` and `%(raw:filters)`

Maybe: s/long/soon/

quoted

for `ref-filter` can be accepted by Git, We can start to let `cat-file` use

s/, We/, we/

quoted

`ref-filter` logic on a large scale! Exciting! But the performance of
`ref-filter`
is still not good. Perhaps I need to find a new breakthrough in the
performance bottleneck of `ref-filter`.

Great, thanks!

Well, there are many typos in this blog... I will correct them in time.

Thanks.
--
ZheNing Hu