git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
* [PATCH 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again
@ 2022-02-07 23:49 Johannes Schindelin via GitGitGadget
  2022-02-07 23:49 ` [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-07 23:49 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin

As reported in https://github.com/git-for-windows/git/issues/3688, calling
git fetch -h outside a repository now results in a very ugly

 BUG: repo-settings.c:23: Cannot add settings for uninitialized repository


The reason is that the prepare_repo_settings() calls (that we introduced to
support sparse index) assume that there is a gitdir, but the hack to allow
parse_options() to handle -h even outside a repository invalidates that
assumption.

One strategy I considered was to move the prepare_repo_settings() calls
after parse_options(). This would work because when parse_options() handles
-h, it exits without returning.

However, this strategy failed in my tests because e.g. cmd_unpack_objects()
does need the pack_use_sparse to be populated correctly before even parsing
the options so that it can be overridden via --sparse/--no-sparse.

Hence the current strategy where the code that prepares the repo settings
and then accesses them is guarded behind the condition that we must have a
gitdir to do so.

Note: There are other instances where prepare_repo_settings() is called
before parse_options(), e.g. in cmd_status(), in seen there are even more
instances (e.g. cmd_checkout_index()). All of those instances that are not
touched by this here patch do have special code to handle -h early, though,
before calling prepare_repo_settings() let alone parse_options().

Johannes Schindelin (2):
  checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  t0012: verify that built-ins handle `-h` even without gitdir

 builtin/checkout.c     | 7 ++++---
 builtin/fetch.c        | 7 +++++--
 builtin/pack-objects.c | 8 +++++---
 builtin/pull.c         | 6 ++++--
 t/t0012-help.sh        | 7 ++++++-
 5 files changed, 24 insertions(+), 11 deletions(-)


base-commit: 4c53a8c20f8984adb226293a3ffd7b88c3f4ac1a
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1139%2Fdscho%2Fprepare_repo_settings-after-parse_options-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1139/dscho/prepare_repo_settings-after-parse_options-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/1139
-- 
gitgitgadget

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  2022-02-07 23:49 [PATCH 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
@ 2022-02-07 23:49 ` Johannes Schindelin via GitGitGadget
  2022-02-08  6:50   ` Junio C Hamano
  2022-02-07 23:49 ` [PATCH 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget
  2022-02-08 11:21 ` [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
  2 siblings, 1 reply; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-07 23:49 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin, Johannes Schindelin

From: Johannes Schindelin <johannes.schindelin@gmx.de>

When we taught these commands about the sparse index, we did not account
for the fact that the `cmd_*()` functions _can_ be called without a
gitdir, namely when `-h` is passed to show the usage.

A plausible approach to address this is to move the
`prepare_repo_settings()` calls right after the `parse_options()` calls:
The latter will never return when it handles `-h`, and therefore it is
safe to assume that we have a `gitdir` at that point, as long as the
built-in is marked with the `RUN_SETUP` flag.

However, it is unfortunately not that simple. In `cmd_pack_objects()`,
for example, the repo settings need to be fully populated so that the
command-line options `--sparse`/`--no-sparse` can override them, not the
other way round.

Therefore, we choose to imitate the strategy taken in `cmd_diff()`,
where we simply do not bother to prepare and initialize the repo
settings unless we have a `gitdir`.

This fixes https://github.com/git-for-windows/git/issues/3688

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 builtin/checkout.c     | 7 ++++---
 builtin/fetch.c        | 7 +++++--
 builtin/pack-objects.c | 8 +++++---
 builtin/pull.c         | 6 ++++--
 4 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/builtin/checkout.c b/builtin/checkout.c
index cc804ba8e1e..1c13d7fedb3 100644
--- a/builtin/checkout.c
+++ b/builtin/checkout.c
@@ -1602,9 +1602,10 @@ static int checkout_main(int argc, const char **argv, const char *prefix,
 	opts->show_progress = -1;
 
 	git_config(git_checkout_config, opts);
-
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	opts->track = BRANCH_TRACK_UNSPECIFIED;
 
diff --git a/builtin/fetch.c b/builtin/fetch.c
index 5f06b21f8e9..e0a05de10ee 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -2014,11 +2014,14 @@ int cmd_fetch(int argc, const char **argv, const char *prefix)
 	}
 
 	git_config(git_fetch_config, NULL);
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	argc = parse_options(argc, argv, prefix,
 			     builtin_fetch_options, builtin_fetch_usage, 0);
+
 	if (recurse_submodules != RECURSE_SUBMODULES_OFF) {
 		int *sfjc = submodule_fetch_jobs_config == -1
 			    ? &submodule_fetch_jobs_config : NULL;
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index ba2006f2212..87cb7b45c37 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -3976,9 +3976,11 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 	read_replace_refs = 0;
 
 	sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1);
-	prepare_repo_settings(the_repository);
-	if (sparse < 0)
-		sparse = the_repository->settings.pack_use_sparse;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		if (sparse < 0)
+			sparse = the_repository->settings.pack_use_sparse;
+	}
 
 	reset_pack_idx_option(&pack_idx_opts);
 	git_config(git_pack_config, NULL);
diff --git a/builtin/pull.c b/builtin/pull.c
index 100cbf9fb85..d15007d93f3 100644
--- a/builtin/pull.c
+++ b/builtin/pull.c
@@ -994,8 +994,10 @@ int cmd_pull(int argc, const char **argv, const char *prefix)
 		set_reflog_message(argc, argv);
 
 	git_config(git_pull_config, NULL);
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	argc = parse_options(argc, argv, prefix, pull_options, pull_usage, 0);
 
-- 
gitgitgadget


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 2/2] t0012: verify that built-ins handle `-h` even without gitdir
  2022-02-07 23:49 [PATCH 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
  2022-02-07 23:49 ` [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
@ 2022-02-07 23:49 ` Johannes Schindelin via GitGitGadget
  2022-02-08 11:21 ` [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
  2 siblings, 0 replies; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-07 23:49 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin, Johannes Schindelin

From: Johannes Schindelin <johannes.schindelin@gmx.de>

We just fixed a class of recently introduced bugs where calling, say,
`git fetch -h` outside a repository would not show the usage but instead
show an ugly `BUG` message.

Let's verify that this does not regress anymore.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 t/t0012-help.sh | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/t/t0012-help.sh b/t/t0012-help.sh
index 91b68c74a15..cbd725ccac8 100755
--- a/t/t0012-help.sh
+++ b/t/t0012-help.sh
@@ -139,13 +139,18 @@ test_expect_success 'git help --config-sections-for-completion' '
 '
 
 test_expect_success 'generate builtin list' '
+	mkdir -p sub &&
 	git --list-cmds=builtins >builtins
 '
 
 while read builtin
 do
 	test_expect_success "$builtin can handle -h" '
-		test_expect_code 129 git $builtin -h >output 2>&1 &&
+		(
+			GIT_CEILING_DIRECTORIES=$(pwd) &&
+			export GIT_CEILING_DIRECTORIES &&
+			test_expect_code 129 git -C sub $builtin -h >output 2>&1
+		) &&
 		test_i18ngrep usage output
 	'
 done <builtins
-- 
gitgitgadget

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  2022-02-07 23:49 ` [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
@ 2022-02-08  6:50   ` Junio C Hamano
  2022-02-08 10:29     ` Johannes Schindelin
  0 siblings, 1 reply; 8+ messages in thread
From: Junio C Hamano @ 2022-02-08  6:50 UTC (permalink / raw)
  To: Johannes Schindelin via GitGitGadget; +Cc: git, Johannes Schindelin

"Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com>
writes:

> From: Johannes Schindelin <johannes.schindelin@gmx.de>
>
> When we taught these commands about the sparse index, we did not account
> for the fact that the `cmd_*()` functions _can_ be called without a
> gitdir, namely when `-h` is passed to show the usage.
>
> A plausible approach to address this is to move the
> `prepare_repo_settings()` calls right after the `parse_options()` calls:
> The latter will never return when it handles `-h`, and therefore it is
> safe to assume that we have a `gitdir` at that point, as long as the
> built-in is marked with the `RUN_SETUP` flag.
>
> However, it is unfortunately not that simple. In `cmd_pack_objects()`,
> for example, the repo settings need to be fully populated so that the
> command-line options `--sparse`/`--no-sparse` can override them, not the
> other way round.
>
> Therefore, we choose to imitate the strategy taken in `cmd_diff()`,
> where we simply do not bother to prepare and initialize the repo
> settings unless we have a `gitdir`.
>
> This fixes https://github.com/git-for-windows/git/issues/3688
>
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> ---

These "prepare" calls are always tricky.  This somehow reminds me of
44c7e62e (repo-settings: prepare_repo_settings only in git repos,
2021-12-06) and e5b17bda (git: ensure correct git directory setup
with -h, 2021-12-06), which were part of the ld/sparse-diff-blame
topic that was merged at 8d2c3732 during the last cycle.

> diff --git a/builtin/checkout.c b/builtin/checkout.c
> index cc804ba8e1e..1c13d7fedb3 100644
> --- a/builtin/checkout.c
> +++ b/builtin/checkout.c
> @@ -1602,9 +1602,10 @@ static int checkout_main(int argc, const char **argv, const char *prefix,
>  	opts->show_progress = -1;
>  
>  	git_config(git_checkout_config, opts);
> -
> -	prepare_repo_settings(the_repository);
> -	the_repository->settings.command_requires_full_index = 0;
> +	if (the_repository->gitdir) {
> +		prepare_repo_settings(the_repository);
> +		the_repository->settings.command_requires_full_index = 0;
> +	}

Looks quite straight-forward and sensible.  Of course, the checkout
command will go berserk and do nonsense if it is allowed to proceed
much further from here without .gitdir, but we know that 

 (1) the only case where .gitdir is NULL at this point is when
     running "-h" from outside a repository (we would have done a
     full RUN_SETUP without demoting it to RUN_SETUP_GENTLY if we
     weren't doing "-h"), and that

 (2) this is soon followed by parse_options() where it notices "-h"
     and stops.

so skipping "prepare" outside a repository is totally safe.

> diff --git a/builtin/fetch.c b/builtin/fetch.c
> index 5f06b21f8e9..e0a05de10ee 100644
> --- a/builtin/fetch.c
> +++ b/builtin/fetch.c
> @@ -2014,11 +2014,14 @@ int cmd_fetch(int argc, const char **argv, const char *prefix)
>  	}
>  
>  	git_config(git_fetch_config, NULL);
> -	prepare_repo_settings(the_repository);
> -	the_repository->settings.command_requires_full_index = 0;
> +	if (the_repository->gitdir) {
> +		prepare_repo_settings(the_repository);
> +		the_repository->settings.command_requires_full_index = 0;
> +	}

Ditto.

>  	argc = parse_options(argc, argv, prefix,
>  			     builtin_fetch_options, builtin_fetch_usage, 0);
> +
>  	if (recurse_submodules != RECURSE_SUBMODULES_OFF) {

Unrelated blank line?

>  		int *sfjc = submodule_fetch_jobs_config == -1
>  			    ? &submodule_fetch_jobs_config : NULL;
> diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
> index ba2006f2212..87cb7b45c37 100644
> --- a/builtin/pack-objects.c
> +++ b/builtin/pack-objects.c
> @@ -3976,9 +3976,11 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
>  	read_replace_refs = 0;
>  
>  	sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1);
> -	prepare_repo_settings(the_repository);
> -	if (sparse < 0)
> -		sparse = the_repository->settings.pack_use_sparse;
> +	if (the_repository->gitdir) {
> +		prepare_repo_settings(the_repository);
> +		if (sparse < 0)
> +			sparse = the_repository->settings.pack_use_sparse;
> +	}
>  
>  	reset_pack_idx_option(&pack_idx_opts);
>  	git_config(git_pack_config, NULL);

This has a bit more sequence of calls until it reaches parse_options(),
but their side effect, when run outside a repository, are all benign
in-core effects, so this is safe, too.

> diff --git a/builtin/pull.c b/builtin/pull.c
> index 100cbf9fb85..d15007d93f3 100644
> --- a/builtin/pull.c
> +++ b/builtin/pull.c
> @@ -994,8 +994,10 @@ int cmd_pull(int argc, const char **argv, const char *prefix)
>  		set_reflog_message(argc, argv);
>  
>  	git_config(git_pull_config, NULL);
> -	prepare_repo_settings(the_repository);
> -	the_repository->settings.command_requires_full_index = 0;
> +	if (the_repository->gitdir) {
> +		prepare_repo_settings(the_repository);
> +		the_repository->settings.command_requires_full_index = 0;
> +	}
>  
>  	argc = parse_options(argc, argv, prefix, pull_options, pull_usage, 0);

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  2022-02-08  6:50   ` Junio C Hamano
@ 2022-02-08 10:29     ` Johannes Schindelin
  0 siblings, 0 replies; 8+ messages in thread
From: Johannes Schindelin @ 2022-02-08 10:29 UTC (permalink / raw)
  To: Junio C Hamano; +Cc: Johannes Schindelin via GitGitGadget, git

Hi Junio,

On Mon, 7 Feb 2022, Junio C Hamano wrote:

> "Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com>
> writes:
>
> > From: Johannes Schindelin <johannes.schindelin@gmx.de>
> >
> > When we taught these commands about the sparse index, we did not account
> > for the fact that the `cmd_*()` functions _can_ be called without a
> > gitdir, namely when `-h` is passed to show the usage.
> >
> > A plausible approach to address this is to move the
> > `prepare_repo_settings()` calls right after the `parse_options()` calls:
> > The latter will never return when it handles `-h`, and therefore it is
> > safe to assume that we have a `gitdir` at that point, as long as the
> > built-in is marked with the `RUN_SETUP` flag.
> >
> > However, it is unfortunately not that simple. In `cmd_pack_objects()`,
> > for example, the repo settings need to be fully populated so that the
> > command-line options `--sparse`/`--no-sparse` can override them, not the
> > other way round.
> >
> > Therefore, we choose to imitate the strategy taken in `cmd_diff()`,
> > where we simply do not bother to prepare and initialize the repo
> > settings unless we have a `gitdir`.
> >
> > This fixes https://github.com/git-for-windows/git/issues/3688
> >
> > Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> > ---
>
> These "prepare" calls are always tricky.  This somehow reminds me of
> 44c7e62e (repo-settings: prepare_repo_settings only in git repos,
> 2021-12-06) and e5b17bda (git: ensure correct git directory setup
> with -h, 2021-12-06), which were part of the ld/sparse-diff-blame
> topic that was merged at 8d2c3732 during the last cycle.

Yes, and I did indeed imitate the way Lessley changed `cmd_diff()`. It
turned out to be much smarter than the first approach I tried (and that I
described as a "plausible approach" in the commit message).

> > diff --git a/builtin/checkout.c b/builtin/checkout.c
> > index cc804ba8e1e..1c13d7fedb3 100644
> > --- a/builtin/checkout.c
> > +++ b/builtin/checkout.c
> > @@ -1602,9 +1602,10 @@ static int checkout_main(int argc, const char **argv, const char *prefix,
> >  	opts->show_progress = -1;
> >
> >  	git_config(git_checkout_config, opts);
> > -
> > -	prepare_repo_settings(the_repository);
> > -	the_repository->settings.command_requires_full_index = 0;
> > +	if (the_repository->gitdir) {
> > +		prepare_repo_settings(the_repository);
> > +		the_repository->settings.command_requires_full_index = 0;
> > +	}
>
> Looks quite straight-forward and sensible.  Of course, the checkout
> command will go berserk and do nonsense if it is allowed to proceed
> much further from here without .gitdir, but we know that
>
>  (1) the only case where .gitdir is NULL at this point is when
>      running "-h" from outside a repository (we would have done a
>      full RUN_SETUP without demoting it to RUN_SETUP_GENTLY if we
>      weren't doing "-h"), and that
>
>  (2) this is soon followed by parse_options() where it notices "-h"
>      and stops.
>
> so skipping "prepare" outside a repository is totally safe.

Indeed.

Plus, do not forget how the second patch in this series ensures that there
are no other calls that would get the hiccups when `gitdir` is `NULL`.

> > diff --git a/builtin/fetch.c b/builtin/fetch.c
> > index 5f06b21f8e9..e0a05de10ee 100644
> > --- a/builtin/fetch.c
> > +++ b/builtin/fetch.c
> > @@ -2014,11 +2014,14 @@ int cmd_fetch(int argc, const char **argv, const char *prefix)
> >  	}
> >
> >  	git_config(git_fetch_config, NULL);
> > -	prepare_repo_settings(the_repository);
> > -	the_repository->settings.command_requires_full_index = 0;
> > +	if (the_repository->gitdir) {
> > +		prepare_repo_settings(the_repository);
> > +		the_repository->settings.command_requires_full_index = 0;
> > +	}
>
> Ditto.
>
> >  	argc = parse_options(argc, argv, prefix,
> >  			     builtin_fetch_options, builtin_fetch_usage, 0);
> > +
> >  	if (recurse_submodules != RECURSE_SUBMODULES_OFF) {
>
> Unrelated blank line?

Whoops, yes. I did mention that I first tried to move the call, right?
That's a left-over from not _completely_ cleaning up after that failed
attempt.

>
> >  		int *sfjc = submodule_fetch_jobs_config == -1
> >  			    ? &submodule_fetch_jobs_config : NULL;
> > diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
> > index ba2006f2212..87cb7b45c37 100644
> > --- a/builtin/pack-objects.c
> > +++ b/builtin/pack-objects.c
> > @@ -3976,9 +3976,11 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
> >  	read_replace_refs = 0;
> >
> >  	sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1);
> > -	prepare_repo_settings(the_repository);
> > -	if (sparse < 0)
> > -		sparse = the_repository->settings.pack_use_sparse;
> > +	if (the_repository->gitdir) {
> > +		prepare_repo_settings(the_repository);
> > +		if (sparse < 0)
> > +			sparse = the_repository->settings.pack_use_sparse;
> > +	}
> >
> >  	reset_pack_idx_option(&pack_idx_opts);
> >  	git_config(git_pack_config, NULL);
>
> This has a bit more sequence of calls until it reaches parse_options(),
> but their side effect, when run outside a repository, are all benign
> in-core effects, so this is safe, too.

And once again, I would like to point out that the regression test is
crucial to keep this benign.

I'll send out v2 shortly.

Ciao,
Dscho

>
> > diff --git a/builtin/pull.c b/builtin/pull.c
> > index 100cbf9fb85..d15007d93f3 100644
> > --- a/builtin/pull.c
> > +++ b/builtin/pull.c
> > @@ -994,8 +994,10 @@ int cmd_pull(int argc, const char **argv, const char *prefix)
> >  		set_reflog_message(argc, argv);
> >
> >  	git_config(git_pull_config, NULL);
> > -	prepare_repo_settings(the_repository);
> > -	the_repository->settings.command_requires_full_index = 0;
> > +	if (the_repository->gitdir) {
> > +		prepare_repo_settings(the_repository);
> > +		the_repository->settings.command_requires_full_index = 0;
> > +	}
> >
> >  	argc = parse_options(argc, argv, prefix, pull_options, pull_usage, 0);
>

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again
  2022-02-07 23:49 [PATCH 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
  2022-02-07 23:49 ` [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
  2022-02-07 23:49 ` [PATCH 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget
@ 2022-02-08 11:21 ` Johannes Schindelin via GitGitGadget
  2022-02-08 11:21   ` [PATCH v2 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
  2022-02-08 11:21   ` [PATCH v2 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget
  2 siblings, 2 replies; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-08 11:21 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin

As reported in https://github.com/git-for-windows/git/issues/3688, calling
git fetch -h outside a repository now results in a very ugly

 BUG: repo-settings.c:23: Cannot add settings for uninitialized repository


The reason is that the prepare_repo_settings() calls (that we introduced to
support sparse index) assume that there is a gitdir, but the hack to allow
parse_options() to handle -h even outside a repository invalidates that
assumption.

One strategy I considered was to move the prepare_repo_settings() calls
after parse_options(). This would work because when parse_options() handles
-h, it exits without returning.

However, this strategy failed in my tests because e.g. cmd_unpack_objects()
does need the pack_use_sparse to be populated correctly before even parsing
the options so that it can be overridden via --sparse/--no-sparse.

Hence the current strategy where the code that prepares the repo settings
and then accesses them is guarded behind the condition that we must have a
gitdir to do so.

Note: There are other instances where prepare_repo_settings() is called
before parse_options(), e.g. in cmd_status(), in seen there are even more
instances (e.g. cmd_checkout_index()). All of those instances that are not
touched by this here patch do have special code to handle -h early, though,
before calling prepare_repo_settings() let alone parse_options().

Johannes Schindelin (2):
  checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  t0012: verify that built-ins handle `-h` even without gitdir

 builtin/checkout.c     | 7 ++++---
 builtin/fetch.c        | 6 ++++--
 builtin/pack-objects.c | 8 +++++---
 builtin/pull.c         | 6 ++++--
 t/t0012-help.sh        | 7 ++++++-
 5 files changed, 23 insertions(+), 11 deletions(-)


base-commit: 4c53a8c20f8984adb226293a3ffd7b88c3f4ac1a
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1139%2Fdscho%2Fprepare_repo_settings-after-parse_options-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1139/dscho/prepare_repo_settings-after-parse_options-v2
Pull-Request: https://github.com/gitgitgadget/git/pull/1139

Range-diff vs v1:

 1:  25d1a2963f2 ! 1:  6e9cdd10a70 checkout/fetch/pull/pack-objects: allow `-h` outside a repository
     @@ builtin/fetch.c: int cmd_fetch(int argc, const char **argv, const char *prefix)
       
       	argc = parse_options(argc, argv, prefix,
       			     builtin_fetch_options, builtin_fetch_usage, 0);
     -+
     - 	if (recurse_submodules != RECURSE_SUBMODULES_OFF) {
     - 		int *sfjc = submodule_fetch_jobs_config == -1
     - 			    ? &submodule_fetch_jobs_config : NULL;
      
       ## builtin/pack-objects.c ##
      @@ builtin/pack-objects.c: int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 2:  49977ff9f4e = 2:  93df4a73dab t0012: verify that built-ins handle `-h` even without gitdir

-- 
gitgitgadget

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH v2 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository
  2022-02-08 11:21 ` [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
@ 2022-02-08 11:21   ` Johannes Schindelin via GitGitGadget
  2022-02-08 11:21   ` [PATCH v2 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget
  1 sibling, 0 replies; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-08 11:21 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin, Johannes Schindelin

From: Johannes Schindelin <johannes.schindelin@gmx.de>

When we taught these commands about the sparse index, we did not account
for the fact that the `cmd_*()` functions _can_ be called without a
gitdir, namely when `-h` is passed to show the usage.

A plausible approach to address this is to move the
`prepare_repo_settings()` calls right after the `parse_options()` calls:
The latter will never return when it handles `-h`, and therefore it is
safe to assume that we have a `gitdir` at that point, as long as the
built-in is marked with the `RUN_SETUP` flag.

However, it is unfortunately not that simple. In `cmd_pack_objects()`,
for example, the repo settings need to be fully populated so that the
command-line options `--sparse`/`--no-sparse` can override them, not the
other way round.

Therefore, we choose to imitate the strategy taken in `cmd_diff()`,
where we simply do not bother to prepare and initialize the repo
settings unless we have a `gitdir`.

This fixes https://github.com/git-for-windows/git/issues/3688

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 builtin/checkout.c     | 7 ++++---
 builtin/fetch.c        | 6 ++++--
 builtin/pack-objects.c | 8 +++++---
 builtin/pull.c         | 6 ++++--
 4 files changed, 17 insertions(+), 10 deletions(-)

diff --git a/builtin/checkout.c b/builtin/checkout.c
index cc804ba8e1e..1c13d7fedb3 100644
--- a/builtin/checkout.c
+++ b/builtin/checkout.c
@@ -1602,9 +1602,10 @@ static int checkout_main(int argc, const char **argv, const char *prefix,
 	opts->show_progress = -1;
 
 	git_config(git_checkout_config, opts);
-
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	opts->track = BRANCH_TRACK_UNSPECIFIED;
 
diff --git a/builtin/fetch.c b/builtin/fetch.c
index 5f06b21f8e9..c8ca4262de6 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -2014,8 +2014,10 @@ int cmd_fetch(int argc, const char **argv, const char *prefix)
 	}
 
 	git_config(git_fetch_config, NULL);
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	argc = parse_options(argc, argv, prefix,
 			     builtin_fetch_options, builtin_fetch_usage, 0);
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index ba2006f2212..87cb7b45c37 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -3976,9 +3976,11 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 	read_replace_refs = 0;
 
 	sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1);
-	prepare_repo_settings(the_repository);
-	if (sparse < 0)
-		sparse = the_repository->settings.pack_use_sparse;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		if (sparse < 0)
+			sparse = the_repository->settings.pack_use_sparse;
+	}
 
 	reset_pack_idx_option(&pack_idx_opts);
 	git_config(git_pack_config, NULL);
diff --git a/builtin/pull.c b/builtin/pull.c
index 100cbf9fb85..d15007d93f3 100644
--- a/builtin/pull.c
+++ b/builtin/pull.c
@@ -994,8 +994,10 @@ int cmd_pull(int argc, const char **argv, const char *prefix)
 		set_reflog_message(argc, argv);
 
 	git_config(git_pull_config, NULL);
-	prepare_repo_settings(the_repository);
-	the_repository->settings.command_requires_full_index = 0;
+	if (the_repository->gitdir) {
+		prepare_repo_settings(the_repository);
+		the_repository->settings.command_requires_full_index = 0;
+	}
 
 	argc = parse_options(argc, argv, prefix, pull_options, pull_usage, 0);
 
-- 
gitgitgadget


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH v2 2/2] t0012: verify that built-ins handle `-h` even without gitdir
  2022-02-08 11:21 ` [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
  2022-02-08 11:21   ` [PATCH v2 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
@ 2022-02-08 11:21   ` Johannes Schindelin via GitGitGadget
  1 sibling, 0 replies; 8+ messages in thread
From: Johannes Schindelin via GitGitGadget @ 2022-02-08 11:21 UTC (permalink / raw)
  To: git; +Cc: Johannes Schindelin, Johannes Schindelin

From: Johannes Schindelin <johannes.schindelin@gmx.de>

We just fixed a class of recently introduced bugs where calling, say,
`git fetch -h` outside a repository would not show the usage but instead
show an ugly `BUG` message.

Let's verify that this does not regress anymore.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 t/t0012-help.sh | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/t/t0012-help.sh b/t/t0012-help.sh
index 91b68c74a15..cbd725ccac8 100755
--- a/t/t0012-help.sh
+++ b/t/t0012-help.sh
@@ -139,13 +139,18 @@ test_expect_success 'git help --config-sections-for-completion' '
 '
 
 test_expect_success 'generate builtin list' '
+	mkdir -p sub &&
 	git --list-cmds=builtins >builtins
 '
 
 while read builtin
 do
 	test_expect_success "$builtin can handle -h" '
-		test_expect_code 129 git $builtin -h >output 2>&1 &&
+		(
+			GIT_CEILING_DIRECTORIES=$(pwd) &&
+			export GIT_CEILING_DIRECTORIES &&
+			test_expect_code 129 git -C sub $builtin -h >output 2>&1
+		) &&
 		test_i18ngrep usage output
 	'
 done <builtins
-- 
gitgitgadget

^ permalink raw reply related	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2022-02-08 11:32 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-02-07 23:49 [PATCH 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
2022-02-07 23:49 ` [PATCH 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
2022-02-08  6:50   ` Junio C Hamano
2022-02-08 10:29     ` Johannes Schindelin
2022-02-07 23:49 ` [PATCH 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget
2022-02-08 11:21 ` [PATCH v2 0/2] checkout/fetch/pull/pack-objects: allow -h outside a repository again Johannes Schindelin via GitGitGadget
2022-02-08 11:21   ` [PATCH v2 1/2] checkout/fetch/pull/pack-objects: allow `-h` outside a repository Johannes Schindelin via GitGitGadget
2022-02-08 11:21   ` [PATCH v2 2/2] t0012: verify that built-ins handle `-h` even without gitdir Johannes Schindelin via GitGitGadget

Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).