Commits


add a gotd session process, split off from the parent process The new session process is able to manipulate files in the repository and keeps track of the read/write client session state. The parent process now restricts its view of the filesystem to the absolute path stored in argv[0], and combines this with unveil "x" on this path. As a result the parent process can only re-exec itself. small tweaks + ok op@


regress: strip trailing / from GOT_TEST_ROOT otherwise most test fails because of double "/" ending in paths and tricking some check into thinking that's failed. ok jamsek stsp


regress: use `local' for some variables


add regress for 'got diff -d' and minor output fix Ensure an actual file path is displayed in the 'got diff -d -c commit path ...' case when one of the specified paths is a deleted file. Prior to this, the unhelpful "/dev/null" label was shown instead. Includes some copypasta fixes noted by op. ok stsp@ op@


got: minor refactor of got_pathlist_free() API Accept flag parameter to optionally specify which pointers to free. This saves callers looping through the list to free pointers. ok + fix stsp@


fix uninitialised fildes variables in libexec helpers Reviewed and uncovered as part of the diff in the forthcoming commit (pathlist API refactor). ok stsp@


got: expand diffstat -d option to 'got diff' Like got log -d, add the switch to 'got diff' to display a diffstat of changes; as per stsp's suggestion, we always include diff output with the diffstat. Disabled in gotwebd (for now :) Includes tweaks from op@ ok stsp@


tweak a gotd regress test to cover incremental 'got fetch' Previously the gotd tests were only running fresh clones.


fix tempfile handling in gotd's got_object_raw_open() There was a bug where we reused a file stored in *outfd from a previous call, resulting in a raw object backed by a file but with tempfile_idx -1. This then caused bad confusion during deltification. Fix this by passing tempfd to read_raw() functions and only setting *outfd in case it is actually required, and returning tempfd to the repository tempfile pool otherwise. ok tracey


gotwebd: templateify gotweb_render_tag ok tracey@


gotwebd: templateify gotweb_render_tags ok tracey@


use a caller-specified size limit for mapped files in got_object_raw_alloc() Without this we end up being confused about whether a raw object has been mapped into memory, leading to crashes. ok op@


adjust a misleading error message in got_object_raw_alloc() ok op@


fix *size not being passed out of read_packed_object_raw() ok op@


tog: add mutt-like =/* keymaps as home/end aliases Also, separate g/G from home/end in the manual and runtime help text as only the former accept a prefixed count modifier. Based on initial diff from Mikhail. ok stsp@


util/got-build-regress.sh: simplify lockfile cleanup trap on EXIT too so the lockfile is deleted on quit (both successful and not.) While here simplify a cd too. ok stsp@


util/got-build-regress.sh: switch from getopt(1) to getopts ok stsp@


show gotd process names in stderr debug logs, too; it is getting too confusing


call realpath() during early startup in gotd's parse.y This ensures that all repositories exist when the process is first started. It will also help to avoid an "rpath" pledge promise in a future gotd which uses a separate session process, by avoiding realpath() calls while starting new processes.


do not treat nonexistent repositories as a fatal error in gotd


fix 'got rebase' not detecting an out-of-date work tree in some cases ok jamsek, op


in got.1, update list of commands which refuse to run during rebase/histedit


show from: and via: headers on consecutive lines in 'got log' and 'tog log' ok jamsek


add -d switch to log usage and sort lexicographically patch by Josiah Frentsos


regress: replace echo \n with printf and heredoc echo \n is not really portable. Not that it matters since the regress suite is already using non-POSIX features (like `local'). However, it's the only place using it and easy to fix. ok jamsek