# [PATCH 0/1] [v2.24.0-rc0 BUG] index.version is not parsed correctly

3 messages from 2019-10-23 to 2019-10-24. Participants: Derrick Stolee via GitGitGadget, Philip Oakley.
Thread: https://gitlist.dev/t/52104

## Derrick Stolee via GitGitGadget, 2019-10-23 20:38

Subject: [PATCH 0/1] [v2.24.0-rc0 BUG] index.version is not parsed correctly
Message-ID: <pull.420.git.1571863137.gitgitgadget@gmail.com>
URL: https://gitlist.dev/e/pull.420.git.1571863137.gitgitgadget%40gmail.com

```
Here is another bug related to changes I made in this release. This time
it's not due to a strange side-effect of submodules but rather that I was
careless while copy-pasting some boilerplate code.

Basically, most of the config options used in repo-settings.c are boolean
settings. Except that "index.version" is an int setting!

I found this while incorporating v2.24.0-rc0 into our microsoft/git fork and
consuming it in VFS for Git. We really care about the index having version
4!

The tests I added around how the index version is set were not adequate to
cover the case of only "index.version=4". Add a test to prevent this issue
from happening again.

Thanks, -Stolee

Derrick Stolee (1):
  repo-settings: read an int for index.version

 repo-settings.c  | 2 +-
 t/t1600-index.sh | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)


base-commit: d966095db01190a2196e31195ea6fa0c722aa732
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-420%2Fderrickstolee%2Frepo-settings-fix-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-420/derrickstolee/repo-settings-fix-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/420
-- 
gitgitgadget

```

## Derrick Stolee via GitGitGadget, 2019-10-23 20:38

Subject: [PATCH 1/1] repo-settings: read an int for index.version
Message-ID: <095b43f606a352527c9e0551adbd8a6af2061176.1571863137.git.gitgitgadget@gmail.com>
URL: https://gitlist.dev/e/095b43f606a352527c9e0551adbd8a6af2061176.1571863137.git.gitgitgadget%40gmail.com
In-Reply-To: <pull.420.git.1571863137.gitgitgadget@gmail.com>

```
From: Derrick Stolee <dstolee@microsoft.com>

Several config options were combined into a repo_settings struct in
ds/feature-macros, including a move of the "index.version" config
setting in 7211b9e (repo-settings: consolidate some config settings,
2019-08-13).

Unfortunately, that file looked like a lot of boilerplate and what is
clearly a factor of copy-paste overload, the config setting is parsed
with repo_config_ge_bool() instead of repo_config_get_int(). This means
that a setting "index.version=4" would not register correctly and would
revert to the default version of 3.

I caught this while incorporating v2.24.0-rc0 into the VFS for Git
codebase, where we really care that the index is in version 4.

This was not caught by the codebase because the version checks placed
in t1600-index.sh did not test the "basic" scenario enough. Here, we
modify the test to include these normal settings to not be overridden by
features.manyFiles or GIT_INDEX_VERSION. While the "default" version is
3, this is demoted to version 2 in do_write_index() when not necessary.

Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
---
 repo-settings.c  | 2 +-
 t/t1600-index.sh | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/repo-settings.c b/repo-settings.c
index 05546db98e..a703e407a3 100644
--- a/repo-settings.c
+++ b/repo-settings.c
@@ -22,7 +22,7 @@ void prepare_repo_settings(struct repository *r)
 	UPDATE_DEFAULT_BOOL(r->settings.core_commit_graph, 1);
 	UPDATE_DEFAULT_BOOL(r->settings.gc_write_commit_graph, 1);
 
-	if (!repo_config_get_bool(r, "index.version", &value))
+	if (!repo_config_get_int(r, "index.version", &value))
 		r->settings.index_version = value;
 	if (!repo_config_get_maybe_bool(r, "core.untrackedcache", &value)) {
 		if (value == 0)
diff --git a/t/t1600-index.sh b/t/t1600-index.sh
index c77721b580..b7c31aa86a 100755
--- a/t/t1600-index.sh
+++ b/t/t1600-index.sh
@@ -87,6 +87,10 @@ test_index_version () {
 }
 
 test_expect_success 'index version config precedence' '
+	test_index_version 0 false 0 2 &&
+	test_index_version 2 false 0 2 &&
+	test_index_version 3 false 0 2 &&
+	test_index_version 4 false 0 4 &&
 	test_index_version 2 false 4 4 &&
 	test_index_version 2 true 0 2 &&
 	test_index_version 0 true 0 4 &&
-- 
gitgitgadget

```

## Philip Oakley, 2019-10-24 11:04

Subject: Re: [PATCH 1/1] repo-settings: read an int for index.version
Message-ID: <e32c2694-e86c-0b93-a680-36b367f4258e@iee.email>
URL: https://gitlist.dev/e/e32c2694-e86c-0b93-a680-36b367f4258e%40iee.email
In-Reply-To: <095b43f606a352527c9e0551adbd8a6af2061176.1571863137.git.gitgitgadget@gmail.com>

```
spelling nit.

On 23/10/2019 21:38, Derrick Stolee via GitGitGadget wrote:
> From: Derrick Stolee<dstolee@microsoft.com>
>
> Several config options were combined into a repo_settings struct in
> ds/feature-macros, including a move of the "index.version" config
> setting in 7211b9e (repo-settings: consolidate some config settings,
> 2019-08-13).
>
> Unfortunately, that file looked like a lot of boilerplate and what is
> clearly a factor of copy-paste overload, the config setting is parsed
> with repo_config_ge_bool() instead of repo_config_get_int(). This means

s/_ge_bool()/_get_bool()/
> that a setting "index.version=4" would not register correctly and would
> revert to the default version of 3.
>
> I caught this while incorporating v2.24.0-rc0 into the VFS for Git
> codebase, where we really care that the index is in version 4.
>
> This was not caught by the codebase because the version checks placed
> in t1600-index.sh did not test the "basic" scenario enough. Here, we
> modify the test to include these normal settings to not be overridden by
> features.manyFiles or GIT_INDEX_VERSION. While the "default" version is
> 3, this is demoted to version 2 in do_write_index() when not necessary.
>
> Signed-off-by: Derrick Stolee<dstolee@microsoft.com>
> ---
Philip

```
