git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [Outreachy] fast-import: Use repo_for_each_pack() for duplicate checks

From
Patrick Steinhardt <ps@pks.im>
Date
Nov 9, 2026, 07:35 UTC
Message-ID
<avF3yR-PYA3_Ih3v@pks.im>
In-Reply-To
<CAH=d44co_ZFA8EJwAoFL4=88ZqWzO65_+GbHDZfjG8amR4cHBw@mail.gmail.com>
Hi Joie,
On Fri, Nov 06, 2026 at 06:44:49PM +0100, Joie Joie wrote:
Show 13 quoted lines
> Hi Patrick,
> 
> I was looking through `fast-import.c` and noticed that both
> `store_object()` and `stream_blob()` still loop through `files->dirs`
> to check whether an object is already in a pack.
> 
> I saw that Git already has `repo_for_each_pack()`, which is used in
> places like `count-objects.c` and `pack-objects.c`.
> 
> Would it be worth using that iterator in these two places instead? The
> idea would be to avoid accessing the files backend directly while
> keeping the current packed-object lookup and duplicate-handling
> behavior unchanged.

No. `repo_for_each_pack()` was introduced as a helper function to make the migration towards pluggable object databases a bit easier. But it is ultimately still flawed, as it assumes that the backend uses packfiles in the first place. So the goal is rather to eventually get rid of this macro altogether in favor of higher-level interfaces that can be implemented by arbitrary backends.

For git-fast-import(1) this will be done by making use of transactions, and Karthik has already started taking a look at it, so I've Cc'd him. For other subsystems we'll need to investigate how to do it properly.

Thanks!
Patrick
Previous: Joie JoieNext: Karthik Nayak
Message 2 of 3 in “[Outreachy] fast-import: Use repo_for_each_pack() for duplicate checks”
  1. Joie JoieNov 6, 2026
  2. Patrick SteinhardtNov 9, 2026
  3. Karthik NayakNov 9, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.