Repository navigation
Conversation
dwl clears the keyboard's focus at the lock but moves the pointer's only with the pointer. Until it moved, the window it rested on took a click or a wheel turn under the lock, and a window a button was held down on kept the pointer through dwl's pressed-button rule: its moves and its release reached a zone's window under the lock. locksession now clears the seat's pointer focus and ends a move or resize by hand; the button's own release comes later and goes to no window. The unlock gives the pointer back as before (destroylock's motionnotify). The GUI suite checks both: a click and a wheel turn under the lock with the pointer unmoved on a zone's window, and a button held on it into the lock, then moved and let go; the window must be sent no pointer event after the lock, with the held press reaching it beforehand as the control. vm-drive gains button:, button-down: and button-up: (QMP btn events), with tests, and wlprobe's cursor mode says each leave, motion and wheel event.
While the state partition waits for its passphrase, a program started waits on it: its loader reads /etc, whose upper layer is there. power.sh started cat, sed, grep and mv on every look and sleep(1) every round, so from a wake to the prompt it stalled on its first look, and the state the chrome reads stopped changing. The 3% power-off was already kryptikd's then. A look now reads /sys and the last state with bash's own read, matches with [[ ]], and keeps its results in variables, not command output. The rounds wait in bash's loadable sleep, as the watchdog's do. The state is still renamed into place, but written in place while the partition waits, as mv is a program. The fixture test runs one look with PATH leading nowhere under a stand-in sealed partition: any program started would fail and show.
An origin row's page that answers 200 with no OpenPGP key at all, as an error page does, was refused as published-key-changed and failed CI's Source manifest on a pull request (zlib against madler.net, once, cleared by a re-run). It now counts as not read, as a page that does not answer: --strict still fails it, so main and Distro read every copy, and a pull request's --unread-warns checks the source against the key held. A page that serves another key is still refused.
The USB core keeps every new device awake until its driver allows autosuspend. For a camera that is uvcvideo's probe, and uvcvideo here binds only for a grant, so a built-in camera stayed awake from boot until its first grant, unlike on a stock system. An eudev rule sets power/control to auto for a USB device with a video control interface on a port the hub calls fixed. Webcams keep the core's default, since uvcvideo keeps a few awake that fail when they sleep. The hardware report gains a "camera usb" line per camera device, giving its port, power/control and runtime status, so the unit's own report shows whether the rule applied; hardware.md gains the row that reads it.
A bare revocation, a key gpg cannot parse or one with no valid user ID lists no primary key either, and these stay refused as a changed key: they are something published there, not an error page. The test is an armor header, or a binary packet's first byte with its top bit set.
kryptik-lock drew every lock surface at its logical size with a buffer scale of 1, so at scale 2 or a fraction the compositor enlarged it and its 5x7 lettering came out soft, unevenly so at 1.25 and 1.5. It now binds wp_viewporter and wp_fractional_scale_v1 where the compositor offers them, draws each output's picture at the surface's size times the scale that output's preferred_scale gives, and shows it at the surface's size through a viewport, as havoc's 0003 does for the terminals. The letter size follows the picture's height in the output's own pixels, so the words keep their size on the screen. Without the two protocols it draws as before. The recipe and desktop-compile.sh generate the two protocols' code for it. The GUI suite locks while the second monitor is at scale 2 and finds a buffer twice its surface's size, then types the password to unlock. Only the scale lines of libwayland's log are kept, as the rest would hold the password's keys.
Neither X86_INTEL_PSTATE nor X86_AMD_PSTATE has a Kconfig default, and x86_64_defconfig names neither, so every machine ran acpi-cpufreq under defconfig's userspace governor, which holds the clock where the firmware left it, and power.sh found no energy_performance_preference to set. Both are built in. amd-pstate starts in active mode with its dynamic EPP off, so the kernel does not override what power.sh sets from the charger. A processor neither takes falls back to acpi-cpufreq under schedutil. Both drivers join the critical options, so a fragment that loses one fails.
…hedutil there, so this adds intel_pstate and the dynamic EPP line to its block
A locale's collation may sort 1-11:1.0 between 1-1:1.1 and 1-1:1.2, and comparing with the last device printed then printed port 1's camera twice. The devices printed are kept instead, and the fixture adds a separate infrared camera on port 11 beside the one on port 1.
…read, [[ ]], loadable sleep, and state written in place while sealed
…he lock, as the keyboard does
…e through a viewport, so its lettering is sharp at 1.5 or 2
…enPGP data is unread, not a changed key
…ne's clock scales and takes the energy preference
The guide said many docks fall back to plain USB-C. On a Thunderbolt or USB4 port, as on the ZBook, a Thunderbolt or USB4 dock connects as USB4 instead, and its DisplayPort is a tunnel that only the unbuilt Thunderbolt driver sets up, so its monitors may stay dark as well as its USB ports and Ethernet. The guide now says so, and that such a dock on a plain USB-C port, or the monitor on the laptop, is the way through; status.md's gap says the same, and the hardware row asks which port the dock was on.
Every trial ended with forget and then ensure, and ensure appended the committed slot's entry last in BootOrder. A firmware that lists network boot after the disks put Kryptik behind it, so each boot waited on the network first, and every update undid an order set in the firmware's setup. kryptik-efiboot gains keep SLOT [ESP]: forget, then the entry where Kryptik's first entry stood. It exits 1 when the forgetting failed and 3 when only the entry did, so boot-success and kryptik-recover keep their two messages. A new entry, from ensure or set-next, goes after the last BootOrder entry that does not boot from the network (a MAC, IPv4, IPv6 or URI node), and so never ahead of another system's entry.
wlroots puts a surface at its position times the output's scale, rounded, and sizes it by rounding its far edge. Where the position times the scale is not whole, as for an odd x at 1.5 or one not a multiple of 4 at 1.25, a buffer drawn at the window's size times the scale is shown a pixel off its own size, and pixman scales it, soft and on every redraw, even for a client such as Firefox that draws at the exact scale. dwl's tiles start wherever mfact and the border put them, so most windows were off the grid. resize now moves a window's surface onto its screen's grid, in steps of 2 at 1.5 and 4 at 1.25 or 1.75, and leaves other scales as they are. A tiled window gives up the pixels from its width and height, so it stays inside its tile; a floating one moves, so a drag does not shrink it. dwl-headless.sh checks it: at scale 1 the one window fills the output inside its border, and once kryptik-display has put the output at 1.25 it is configured smaller by the step from its border to the next multiple of 4.
USB4 also runs at 20 Gbps, a 20Gbps mark can be plain USB 3.2, and some makers mark USB4 ports with a DisplayPort or power icon or nothing, so the guide says the laptop's specifications tell which ports are USB4.
… dock on such a port may leave its monitors dark, as no USB4 driver is built
…each window on a whole pixel of its screen, so a buffer drawn at the exact scale is shown at its own size
…e in the boot order its slot's entry had, rather than going last
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The batch after #670, stacked on zbook-batch-5. It is proved on its own Distro run before main moves.
Every member was read by someone other than its author, at the head merged here (see the ledger):
read,[[ ]], the loadablesleep, and state written in place while sealed.--strictstill fails it.#671's and #673's own Source manifest failed only on zstd's key at facebook.github.io, which #663 (in zbook-batch-5) fixes.
Effect on the caches: #676 rebuilds the kernel (stage 05). No stage 04 row changes, so no image rebuilds because of this batch; the images rebuild only because a new branch has no cache of its own.