Keep track of parent commits that will be processed as part of looping over the commit queue provided by the main process, and do not add these commits to the queue again. Fixes pointless traversal of commits on the queue which will simply be skipped. The end result is the same either way. ok tracey

got-read-pack forgot to send a tree-enumeration-done message to the main process if the tree of a given commit had already been traversed. The main process would then not add the corresponding commit to the pack file, even though it should be added. Found while using 'got send' towards gotd in order to populate an empty repository on the server with non-trivial history, where some commits always ended up missing due to this bug. ok tracey

ok jrick

Diff from Mikhail, thanks!

ok op, jamsek

ok stsp@

Reported and tested by dv: rather than print a new line to avoid clobbering the shell prompt when exiting tog with the alternate screen buffer disabled^, adopt naddy's vi(1) solution by looping through visible views to delete the topmost line, which inserts an empty line at the bottom so we don't need to print an empty line to ensure a clean prompt. ^: xterm -xrm 'XTerm*titeInhibit: 1' or run tog in the console ok naddy@

If the diff algorithm, ignore whitespace-only changes, force ASCII text diff, show committer/author, or reference sort order keymaps are toggled, report action to the bottom left of screen. Based on initial patch by Mikhail. ok + tweaks op@

In view_loop(), we initialise at a refresh rate of one tenth per second before dropping to one second updates; the reduction was never hit due to the typo. Reviewed as part of the following diff adding action reports to user events. ok op@

and while here drop the second pledge() call since unveil is already locked by apply_unveil. ok stsp

ok stsp

would have caught the fixed failure in disposing early-closed connections in gotd. ok jamsek, stsp

Export parse_command (now got_serve_parse_command) from lib/serve.c and use it to validate the command line in gotsh instead of just checking that the -c argument starts with 'git-receive-pack' or 'git-upload-pack'. Invalid usage now always fails before opening the socket, while before it wasn't always the case. This also means that invalid usage doesn't count towards the limits. ok jamsek, stsp

Otherwise gotd keeps the (client-closed) sockets around and may prevent new connections from being established since they still count for the limits. ok jamsek, stsp

This purposefully breaks the 'previous' button in the commits and briefs page. It's hard to find the parent of a commit since they can only be iterated forward. The way the previous button was generated was to walk the history from the HEAD down to the specified commit. This is costly but more importantly leads to issue when dealing with paths that were deleted from the repository. Discussed with stsp and tracey, ok jamsek.

It's wasteful since `refs' is already populated before the loop, and released after. It also leaks the content of `refs' the first time the loop is entered. ok jamsek

ok jamsek

Rework them so that they allow to set the Status header (the HTTP status code; only way since we're behind FastCGI) and optionally a Location. Since they're now unused outside of gotweb.c, mark them as static. They also used to always return NULL so the error is pointless; return the -1 on failure though. While here, rename to gotweb_reply and gotweb_reply_file. ok jamsek

ok jamsek