JSON strings are made of UNICODE codepoints, of which only \, " and control characters have to be escaped, and the whole document MUST be encoded in UTF-8. The current code generates invalid strings for non-ASCII characters, so it has to be made UTF-8 aware. tedu' isu8cont() can't be used since it allows surrogate pairs and overlong sequences which will cause decoding errors on the receiving side. Similarly, mbtowc() depends on the current locale and could cause issues in -portable. Instead, bundle Björn Höhrmann's "Flexible and Economical UTF-8 Decoder" and use it to parse the text. Decoding errors results in the replacement character U+FFFD being emitted and the bytes considered so far to be discarded; the decoder is then restarted with the next byte. Git commit messages don't carry the notion of the encoding, but it's reasonable to expect UTF-8 (which is a superset of ASCII). For other more esotic encodings, the commit id can be used to manually extract the data. ok stsp@

long lines are not manegeable. split the json over multiple lines, then use ed to join everything back in a single one.

The http-server script is already validating the headers, doing so in the regress too is not helpful.

At least for now, we don't support following redirects nor retrying the post, so consider a 3XX status a failure too.

ok stsp@

A path of "/" is valid and trailing slashes must be preserved. ok stsp@

Similar to the `got cat' output; it's needed to un-ambiguosly parse the content of the notification, which is already useful to parse the email content and invaluable for the upcoming got-notify-http. ok stsp@

This function begins a breadth-first traversal. The new name makes it easier to distinguish from got_commit_graph_toposort().

This function starts a breadth-first traversal, not necessarily first-parent.

Fixes a problematic case of spurious conflicts encountered by naddy@ on landry's firefox package git repository. The current implementation of toposort is expensive, so this might make rebase appear to run slowly on large repositories. However, this is better than letting users deal with spurious conflicts. ok op@

Because the current implementation of toposort is expensive, add a flag which enables it. I would rather not have this option and just use toposort by default, however more work is required to achieve acceptable performance. ok op@

The algorithm implemented here is based on a description I read on github's blog. See code comments for details. ok op@

The leak is present in got_privsep_recv_traversed_commits. There is an edge case where it receives consecutive imsgs. The first behaves as normal and we got_object_id_dup the last commit id for changed_commit_id. The following imsg(s) then still allocates the last commit id, leaking the one(s) prior allocated. Patch by Kyle Ackerman

Just clear any left-over iteration state and begin a fresh iteration instead of returning GOT_ERR_ITER_BUSY if the caller did not loop through the entire graph. This change currently doesn't matter much since all existing callers only do a single pass over the graph. But it frees up an error code and makes this API more flexible.

This error popped up in op's regress builder: got-fetch-pack: error 118 "test-repo: Permission denied": poll: unexpected end of file ok op@

This makes it easier to tweak the read/write code paths separately.