Repository navigation
v0.8.2
Upgrade if your computer runs systemd 258 or later (Ubuntu 26.04, Fedora 43 and 44, Debian testing, Arch): there, Pots 0.8.0 and 0.8.1 could rewrite ${…} in the bot's commands, so a command ran changed or not at all. 0.8.2 fixes that. A Slack call with a file ID no longer hangs, and Slack requests give up within about 2 minutes.
Fixed
- On systemd 258 and later, a command in a scope could run changed, or not at all. Since 258,
systemd-run --scopeexpands variables in the command it runs by default (--expand-environment): in each bash, screen and routine command,${…}became the value of an environment variable or nothing (${f%.txt},${i}in a loop),$$became$, and a command that began with$, such as$HOME/bin/tool, was left out, so nothing ran and the command succeeded with no output. Ubuntu 26.04, Fedora 43 and 44, Debian testing and Arch ship such a systemd. The command now goes through systemd-run in the environment, which systemd-run passes on as it is, so it reaches bash as the bot wrote it with every systemd version; no new option, so older systemd works as before. call slack files.delete(orfiles.info) with a file ID hung for half an hour. Pots read everyfileparam as the path of a file to upload, so the ID became a body that never ended; Slack answered 408, and the Slack SDK tried again for about 30 minutes. Now only afilethat is a local path (/…or~/…), also infile_uploads, is read, before the call, so a file that is not there fails at once; any otherfilegoes to Slack as it is.- A Slack call, send or reaction that Slack does not answer fails within about 2 minutes. A request that has no answer from Slack 30 seconds after it began fails; one that sends a file gets a minute more for each 5 MB, so a large upload on a slow link still goes through (down to about 0.7 Mbit/s). One that fails before Slack answers (no answer, an HTTP error, too many requests) is tried again until a try fails a minute or more after the first began. The Slack SDK's default tried for about 30 minutes, and held the bot's turn that long. The error says how long Pots tried, for example
An HTTP protocol error occurred: statusCode = 408; tried for 62 s, then gave up. Discord and Telegram calls stay as they were, since they had limits already: Discord's client gives up a try after 15 seconds and tries a failed request at most 3 more times, besides the waits Discord asks for when it limits the rate; Telegram's calls try again only when Telegram asks for a wait of up to a minute, at most twice. - "Pots's desktop" on the desktop page. The page and the password page write the bot's name with the possessive as English does: "Coly's desktop", "Pots' desktop".
- The
desktop-pageskill said that a maximized window follows the phone's shape. It does only when the window can be that narrow: Chrome is at least 500 pixels wide times its scale, 625 at 1.25, more than the 500 of a phone-shaped desktop. Then openbox makes the window as narrow as it can, so it reaches past the edge, and XFCE's xfwm4 leaves it at its old size and stops changing its size until it is maximized again. The skill now says so, and tells the bot to keep its browser at scale 1, zoom sites in instead, and maximize a window again after the link if it needs to. test/scopes.test.tsread a command's file before the command wrote it. The shell makes the file empty beforecatwrites it, so the test could read an empty file; it now waits for the written line.
What's Changed
- 0.8.2 fixes: Slack file-ID hang + request bound; Pots' desktop; desktop-page window sizes; scopes test race; no systemd 258+ expansion in scoped commands by @Crokily in #55
Full Changelog: v0.8.1...v0.8.2