apply load-url-use-heuristic for command line arguments too This semplifies a bit the handling. humanify_url() now takes the base URL which we can then use when not using the heuristics. Command-line arguments now take an implicit base of <file://$PWD/> when the heuristics are disabled, so that foo.gmi resolves to the local file even without <./>. See github issue #10.

iritest: one more test, just in case It stemmed from my misuse of iri_parse(). There's nothing wrong in the parser (that I found at least), but one more test won't harm.

convert humanify_url to iri APIs

convert ir_select_reply() to iri

add missing include

fix iritest CFLAGS; add iritest to TESTS

sync changelog

add the up, root and home commands These navigates respectively one level up in the path hierarchy, to the root directory and to the "home". The home is considered to be the first path component in the form of ~username. `home' is yet to be implemented.

sort

add load-url-use-heuristic knob This is still WIP and may change before the next release. It could be worthwile to resolve the URLs in load-url using the current page as base, so that while browsing <gemini://a.org/b/c>, using load-url to go to "/" would load <gemini://a.org/> instead of <file:///>. The new option `load-url-use-heuristic', if unset, provides exactly this behaviour. It is still a bit rough to use, as for e.g. loading <d.net> would now yield <gemini://a.org/b/d.net> instead of <gemini://d.net>. Will need more work in-tree, but for now add it so we can reason about it. See github issue #10

fmt

don't strlcat in a loop not that of a big deal, but since we memset() at the start, just add the spaces directly instead of using strlcat. saves some cpu cycles.

fix wrapping of link lines

fix comment

add a comment explaining the overridden rule