Add support for SHA-256 repositories #1302
Conversation
src/tree.c
Outdated
| @@ -58,7 +58,7 @@ push_tree_stack_entry(struct view *view, const char *name, struct position *posi | |||
| */ | |||
|
|
|||
| #define SIZEOF_TREE_ATTR \ | |||
| STRING_SIZE("100644 blob f931e1d229c3e185caad4449bf5b66ed72462657\t") | |||
| STRING_SIZE("100644 blob ") + repo.hash_len + 1 | |||
There was a problem hiding this comment.
maybe
| STRING_SIZE("100644 blob ") + repo.hash_len + 1 | |
| STRING_SIZE("100644 blob ") + repo.hash_len + STRING_SIZE("\t") |
| @@ -203,6 +203,8 @@ following variables. | |||
| |%(repo:is-inside-work-tree) | |||
| |Whether Tig is running inside a work tree, | |||
| either `true` or `false`. | |||
| |%(repo:hash_len) |The hash algorithm used for the repository, e.g. `sha1` | |||
| or `sha256`. | |||
There was a problem hiding this comment.
Maybe this should be called %(repo:extensions.objectformat) to use the standard Git term.
Actually, is there even a need to expose this at all?
I don't think we expose any other config options. Users can already consult something like
git --git-dir=%(repo:git-dir) config extensions.objectformat
There was a problem hiding this comment.
Thanks for the review @krobelus. %(repo:hash_len) is actually a by-product of storing hash_len in the repo struct, but it's the best compromise I have found. I'll make it more elegant.
There was a problem hiding this comment.
oh it is exposed by the REPO_INFO macro. Maybe define it outside that
src/options.c
Outdated
| @@ -1531,6 +1531,9 @@ read_repo_config_option(char *name, size_t namelen, char *value, size_t valuelen | |||
| else if (!strcmp(name, "core.abbrev")) | |||
| parse_int(&opt_id_width, value, 0, SIZEOF_REV - 1); | |||
There was a problem hiding this comment.
Not really important: we have two places where we call
parse_int() with a max of SIZEOF_REV - 1.
The effective max should maybe be lower for sha1 repos.
But I'm not sure if that would cause unnecessary friction, so better keep it.
There was a problem hiding this comment.
My initial code used hash_len but we haven't read extension.objectFormat yet here, so we can't use it. I'll improve this part so that the same configuration can be used in global files for both sha1 and sha256 repositories.
| @@ -19,7 +19,7 @@ | |||
| #include "tig/util.h" | |||
|
|
|||
| struct commit { | |||
| char id[SIZEOF_REV]; /* SHA1 ID. */ | |||
| char id[SIZEOF_REV]; /* Hash ID. */ | |||
There was a problem hiding this comment.
this looks good apart from one comment about %(repo:hash_len).
I haven't used sha256, I'll try it if I get around to.
It's a shame we can't reuse existing unit tests without modification.
There was a problem hiding this comment.
Actually git fast-export and git fast-import allow to port the tests without too much effort and also to get a nice sandbox from an existing repository.
koutcher
force-pushed
the
sha-256
branch
4 times, most recently
from
f4c4757 to
6475d09
Compare
koutcher
force-pushed
the
sha-256
branch
2 times, most recently
from
98b589d to
357af1a
Compare
Use variable extensions.objectformat, if available, to determine the length of the hash id in the current repository.
With Git 2.42, SHA-256 repositories are no longer experimental. This PR adds support to Tig for SHA-256 repositories.