Commits


make close(2) failure checks consistent; check 'close() == -1' everywhere ok millert, naddy


switch reflist to TAILQ; insert elements more efficiently for sorted input ok naddy


avoid signed vs unsigned comparisons in fetch.c ssizeof() macro idea courtesy of millert@ ok millert@


use size_t for loop indices to avoid signedness warnings; from emaste@freebsd


use fchmod(2) instead of chmod(2). it is less racy, and we just opened the descriptor. ok naddy@ stsp@


fix permissions mode bits for fetched pack files; patch by Alisdair MacLeod


handle failed connection attempts to git:// servers; patch by jrick


strip trailing slashes from repository path in got_fetch_parse_uri()


remove redundant leading slashes from repository path when parsing URIs


Stop including <sys/syslimits.h> directly. POSIX says the limits defined there are available from <limits.h>, which almost all affected source files already included anyway. ok millert stsp


use the POSIX standard endian functions and explicitly include <endian.h> ok stsp


do not rely on <zlib.h> to pull in <unistd.h> ok stsp


don't override port configuration with explicit argument if the url doesn't contains it it permits to .ssh/config to be honored same diff from stsp@


stop verifying pack file checksum in the main process during clone/fetch Both got-fetch-pack and got-index-pack now verify the pack file checksum. This means we can avoid reading the entire pack file in the main process just to verify its checksum and avoid a noticable stall between fetching and indexing on slow machines.


represent packfile hash as byte array in the privsep layer, not as object ID