Re: [PATCH v6 18/27] files-backend: replace submodule_allowed check in files_downcast()
- From
Michael Haggerty <mhagger@alum.mit.edu>
- Date
- Mar 19, 2017, 21:18 UTC
- Message-ID
- <004082a0-871f-cd41-edc5-522b5c2b3b58@alum.mit.edu>
- In-Reply-To
- <20170318020337.22767-19-pclouds@gmail.com>
On 03/18/2017 03:03 AM, Nguyễn Thái Ngọc Duy wrote:
Show 58 quoted lines
> files-backend.c is unlearning submodules. Instead of having a specific
> check for submodules to see what operation is allowed, files backend
> now takes a set of flags at init. Each operation will check if the
> required flags is present before performing.
>
> For now we have four flags: read, write and odb access. Main ref store
> has all flags, obviously, while submodule stores are read-only and have
> access to odb (*).
>
> The "main" flag stays because many functions in the backend calls
> frontend ones without a ref store, so these functions always target the
> main ref store. Ideally the flag should be gone after ref-store-aware
> api is in place and used by backends.
>
> (*) Submodule code needs for_each_ref. Try take REF_STORE_ODB flag
> out. At least t3404 would fail. The "have access to odb" in submodule is
> a bit hacky since we don't know from he whether add_submodule_odb() has
> been called.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
> refs.c | 15 ++++++---
> refs/files-backend.c | 86 +++++++++++++++++++++++++++++++++-------------------
> refs/refs-internal.h | 9 +++++-
> 3 files changed, 73 insertions(+), 37 deletions(-)
>
> [...]
> diff --git a/refs/files-backend.c b/refs/files-backend.c
> index db335e4ca6..7d8d4dcc16 100644
> --- a/refs/files-backend.c
> +++ b/refs/files-backend.c
> [...]
> @@ -1923,21 +1935,23 @@ static struct ref_iterator *files_ref_iterator_begin(
> struct ref_store *ref_store,
> const char *prefix, unsigned int flags)
> {
> - struct files_ref_store *refs =
> - files_downcast(ref_store, 1, "ref_iterator_begin");
> + struct files_ref_store *refs;
> struct ref_dir *loose_dir, *packed_dir;
> struct ref_iterator *loose_iter, *packed_iter;
> struct files_ref_iterator *iter;
> struct ref_iterator *ref_iterator;
>
> - if (!refs)
> - return empty_ref_iterator_begin();
> -
> if (ref_paranoia < 0)
> ref_paranoia = git_env_bool("GIT_REF_PARANOIA", 0);
> if (ref_paranoia)
> flags |= DO_FOR_EACH_INCLUDE_BROKEN;
>
> + refs = files_downcast(ref_store,
> + REF_STORE_READ | (ref_paranoia ? 0 : REF_STORE_ODB),
> + "ref_iterator_begin");
> + if (!refs)
> + return empty_ref_iterator_begin();
> +I realize that this `if (!refs)` check existed in the old code, but isn't it pointless? If `refs` were NULL, `files_downcast()` would have already segfaulted, I think.
Show 11 quoted lines
> [...] > diff --git a/refs/refs-internal.h b/refs/refs-internal.h > index dfa1817929..0cca280b5c 100644 > --- a/refs/refs-internal.h > +++ b/refs/refs-internal.h > @@ -481,12 +481,19 @@ struct ref_store; > > /* refs backends */ > > +/* ref_store_init flags */ > +#define REF_STORE_READ (1 << 0)
I asked [1] in reply to v5 whether `REF_STORE_READ` is really necessary but I don't think you replied. Surely a reference store that we can't read would be useless? Can't we just assume that any `ref_store` is readable and drop this constant?
Show 14 quoted lines
> +#define REF_STORE_WRITE (1 << 1) /* can perform update operations */ > +#define REF_STORE_ODB (1 << 2) /* has access to object database */ > +#define REF_STORE_MAIN (1 << 3) > + > /* > * Initialize the ref_store for the specified gitdir. These functions > * should call base_ref_store_init() to initialize the shared part of > * the ref_store and to record the ref_store for later lookup. > */ > -typedef struct ref_store *ref_store_init_fn(const char *gitdir); > +typedef struct ref_store *ref_store_init_fn(const char *gitdir, > + unsigned int flags); > > typedef int ref_init_db_fn(struct ref_store *refs, struct strbuf *err);
Michael
[1] http://public-inbox.org/git/8fafd49f-71a6-ee97-6a69-c23e23c5d515@alum.mit.edu/