an opinionated terminal multiplexer in C with programmable pane chrome https://git.sl0p.foo/slosh.git/atom/docs/index.md?h=docs-new 2026-09-02T10:34:31Z 2026-09-02T10:34:31Z Peter Geissler peter@haxx.in 2026-09-02T10:34:31Z urn:sha1:5f5f0aaf5ef3f86b6345d26d0abc9dc37f4acbf3 2026-09-02T10:32:45Z Peter Geissler peter@haxx.in 2026-09-02T10:32:45Z urn:sha1:794e22914666775cab0b1b730a000b2529cec6b5 2026-09-02T09:55:35Z Peter Geissler peter@haxx.in 2026-09-02T09:55:35Z urn:sha1:93e8b6a98de3f9e8f3176199e7974f4bc8a66a1f 2026-08-29T18:25:39Z user user@clank 2026-08-29T18:25:39Z urn:sha1:31c32f9c20e3dd5045fa053fc8e7e606a8139397 2026-08-28T12:22:57Z Peter Geissler peter@haxx.in 2026-08-28T12:22:57Z urn:sha1:7941820413d82b3d62d7fc2ac15ef0e1b3b69ee4 The docs offered a native build as an equal alternative -- "the same command builds natively on the Windows box itself, with zig on PATH and a make" -- and it does not work. Makefile.windows needs sed, cat, mkdir -p and shell grouping, none of which cmd.exe has; and on ARM64 `zig cc` in zig 0.16.0 dies with an access violation on a one-line C file, so the compiler cannot run there at all. The zip that does it matches the published shasum, so this is upstream rather than a bad download. Nobody following those two sentences reached a binary. So they are gone, along with the "# host arch, native" usage line in the makefile and the "cross-compiles too" aside in the two install sections, which both implied that building on the box was the normal path and the cross-compile a bonus. It is the other way round, and now the pages say so once, in the same words each time. `builds and runs natively` on the doc's first line loses `builds`: what is native about slosh.exe is the binary, not where it was made. 2026-08-27T22:23:19Z Peter Geissler peter@haxx.in 2026-08-27T22:11:31Z urn:sha1:d299677c834b7199d47f35817ed36d83c839825b An install subsection alongside macOS and Linux, and a page under Inside for the rest: how to build it, what changes about running it (%ComSpec% panes, sessions in %LOCALAPPDATA%), and the reasoning the code can only gesture at in comments. The three traps are written down in full, because each one presents as something other than what it is and cost real time: a pane with a title and no body (the child inherited slosh's stdout rather than the pseudoconsole), a pane that stays alive after `exit` (a ConPTY's output pipe belongs to the console, not the child, so the read never ends), and an editor refusing a path that plainly exists (cmd.exe and CommandLineToArgvW disagree about what \" means, and separately the CRT accepts a mixed-separator path where a GUI dialog will not). 2026-08-27T14:19:25Z user user@clank 2026-08-27T14:19:25Z urn:sha1:b6fb41cea7610e56280a4db6572bbad955a9b667 'With Homebrew (macOS or Linux)' followed by a source build made every reader parse which half was theirs; now macOS and Linux each have a heading, and you read yours. macOS is the tap; Linux is the same tap under Homebrew on Linux, or the source build, which the docs landing page had never mentioned brew at all. Also fixes the README telling people to run bare 'make', which prints the help text (the default goal) and builds nothing. install.sh had this figured out and commented on it; the README had not. 2026-08-27T14:11:26Z user user@clank 2026-08-27T14:11:26Z urn:sha1:238b1a7eb51c576254537705597f8a039e1d9c86 Some 230 sentences hung their qualifications between dashes, and at that density the tic reads as the writing's signature rather than a tool it occasionally picks up. Each one is now the punctuation the sentence actually wanted: a colon where the aside defines, parentheses where it whispers, a comma where it barely pauses, a full stop where two thoughts were pretending to be one. Definition lists and reference tables take colons throughout. Two double hyphens remain on purpose, both literal: the argv separator in a shell example, and a table cell that means 'takes nothing'. 2026-08-27T14:04:37Z user user@clank 2026-08-27T14:04:37Z urn:sha1:c76159dd380bf5a37d1a24e5295c42c626090c5c The 'Why another one' section was two paragraphs of case-making against other multiplexers, and a landing page that defends its existence reads as unsure of it. What the section said truthfully -- the vendored core, recompute-not-remember, scriptable to its depth -- design.md already says where it belongs. The one-line closer stays: 15k lines of C and no dependencies is a fact, not a defence. 2026-08-27T14:02:11Z user user@clank 2026-08-27T14:02:11Z urn:sha1:d7742f16b81ef09e74c0cbe6999d07100e5ef63c Same reasoning as the README's (24a6db01): the sketch says nothing the first paragraph did not, and on the rendered site it sat where the reader was deciding whether to keep reading.