Thread (238 messages) flat view 238 messages, 6 authors, 2024-10-06

Re: [GSoC][PATCH v15 0/9] ref consistency check infra setup

From: Patrick Steinhardt <hidden>
Date: 2024-08-06 07:32:49

On Tue, Aug 06, 2024 at 12:43:22AM +0800, shejialuo wrote:
Hi All:

This version handles the following problems:

1. Patrick advices that I should not use `va_copy` in the changed
`report` function. Actually this is a mistake, this version avoids
redundant `ap` copy.
2. Patrick advices I should rebase [v14 05/11] into [v14 04/11]. I
follow this advice in this version.
3. Patrick advices that we should put [v14 06/11] before we introduce
ref-related operations. This version reorders the commit sequence. It's
a minor change.
4. Patrick suggests at current we should not add `git refs verify`
command into "git-fsck(1)". This is because we should disable this new
check by default for the users. Many users use "git-fsck(1)" in their daily
workflow. We should not be aggressive. However, if we provide this
mechanism in this series, we will again make more complexity. So this
version drop patch [v14 09/11]. Also because of dropping, change the
test file to use "git refs verify" command instead of "git fsck"
command.
5. Patrick suggests that we should use `ends_with` instead of
`strip_suffix`, fix.

There is another important problem this patch solves:

At v13, Junio has suggested that the `files_fsck_refs_fn` should be
adapted to Patrick's change. Actually, I made a bad design before. I
should always pass the `ref_store` structure. So I change it to

  -typedef int (*files_fsck_refs_fn)(struct fsck_options *o,
  -				  const char *gitdir,
  +typedef int (*files_fsck_refs_fn)(struct ref_store *ref_store,
  +				  struct fsck_options *o,
            const char *refs_check_dir,
            struct dir_iterator *iter);

`gitdir` could be got by using `ref_store` parameter. By using
`ref_store` parameter, we provide extensibility here. If something else
change, we merely need to change "files_fsck_refs_fn" prototype.

Because I drop one patch and rebase one patch. I provide the `interdiff`
for reviewers to make the life easier.

Due to the deadline of the GSoC, I will speed up the review feedback
process.
I've got another small set of nits, almost not worth addressing. I was a
bit torn whether to send them or not as the series is in a good shape
already, in my opinion. But let's maybe wait one or two more days for
additional feedback, and then (hopefully) reroll this a final time.

Thanks for all your work!

Patrick

Attachments

Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help