Commit Briefs
don't create signed tag objects with trailing NUL
Although Git itself did not care, the superfluous NUL at the end of the tag object was breaking GitHub's SSH signature detection. ok stsp@
fix snprintf error handling
follow the "proper secure idiom" described in the CAVEATS section of printf(3). reminded by tb@ and millert@
Prevent memory leak when asprintf fails.
Found by llvm's scan-build. OK stsp
fix pipe usage for linux
This uses the correct read and write ends of the fds returned by pipe(2) on linux. It also moves away from creating FILE* streams with fdopen and reading the stream with buf_load (which performs a fstat, and breaks due to a zero length file size on linux) by instead reading from the fd itself. Reported by abieber@, with assistance from stsp@ ok stsp@
fix tag signing when the key file does not exist
This should fail without creating any tag. Before, ssh-keygen(1) would print an error to stderr, but got would create an unsigned tag. ok op@
create and verify tags signed by SSH keys
This adds a new -s flag to 'got tag' that specifies the signer identity (for example, a key file) of the tagger. The tag object will include a signature that validates each of the tag object headers and the tag message. Verifying these signed tags requires maintaining an allowed signers file which maps signer identities (i.e. the email address of the tagger) to SSH public keys. See ssh-keygen(1) for more details of the allowed signers file. After creating this file and providing the path to it in got.conf(5) using the allowed_signers option, tags may be verified using with 'got tag -V tag_name'. The return code will be non-zero if a signature fails to verify. ok stsp@
build with -Wmissing-prototypes
ok stsp@
inline struct got_object_id in struct got_object_qid
Saves us from doing a malloc/free call for every item on the list. ok op@
add O_CLOEXEC (close-on-exec) flag to open(2) calls
suggested by millert ok thomas_adam
for portability, handle errno variations upon open(2) failure with O_NOFOLLOW
Problem pointed out by naddy for FreeBSD -portable. Discussed with millert, thomas adam, and naddy.
add a 'got merge' command for creating merge commits
Additional testing by Thomas Adam. ok tracey
switch from SIMPLEQ to equivalent STAILQ macros
The singly-linked tail queue macros were added to OpenBSD 6.9 and are more widely available on other systems. ok stsp
add checksum support to got_deflate_to_file()
This will eventually be used by 'gotadmin pack'. Checksum init and finalization will need to be done by the caller since many objects will be written out in compressed form while we are computing checksums across the entire pack file. ok millert, naddy
fix potential type mismatches between format specifiers and arguments
Cast printf arguments of type time_t and off_t to long long to match the %lld format specifier on platforms where this might not be the case. In parse.y, switch the number variable to long long because all its interactions are with that type anyway. ok millert stsp
make 'got histedit' collapse folded add+delete operations into a no-op
If a merged commit wants to delete a locally added file, and this locally added file matches the content which was deleted in the commit being merged, we can go ahead with the deletion because there is no risk of data loss. fixes the histedit problem reported by jrick on freenode
allow creation of commits which carry unmodified submodule tree entries along
approach suggested by ori@ and matches how git9 behaves ok semarie@ (who can now work with Rust-related Git repos containing submodules)