First launch no longer sits on "Connecting..." doing nothing. The first
time TRCC sees a screen it downloads that screen's themes, previews and masks
— about 30 MB for a rectangular panel, in six pieces. It was doing that before
opening the window, so on a slow connection you got a static "Connecting
Thermalright LCD Display..." for minutes with no sign it was working, and on a
connection that stalled, for very much longer. TRCC also asked for a two-minute
limit per download and did not get one: the limit it set never covered looking
the address up, and applied separately to each of the four addresses GitHub
answers with, so a single unreachable download could hold the window shut for
the better part of an hour.
The download now happens in the background. The window opens straight away,
and your themes and masks appear in their grids as they arrive. If a download
fails you get a working app with an empty grid instead of an app that never
opens. (#275)
Images and screen mirroring were cropped on wide panels. On 1600x720,
1280x480, 1920x462 and 854x480 screens, display send-image and screen
mirroring sent a picture the wrong way round for the panel at the two most
common angles, 0 and 180 degrees — the firmware then painted only the part
that overlapped, so roughly a third of the screen lit up. Themes were always
correct; only these two paths were wrong, which is why it looked like
"works rotated, broken straight". Now checked on all ten affected panels at
all four angles. If it still looks wrong on the glass, say so — the last
unconfirmed detail is which way round it lands on a real screen. (#262, #203)
TRCC also warns in the log when it is about to send a picture whose shape
disagrees with what it told the panel, instead of reporting success and
lighting up a third of the display.
Unplugging and replugging a screen no longer leaves it dead. If the
cooler re-enumerated messily on reconnect — a couple of failed attempts before
it settled — TRCC kept talking to the handle from the first attempt. The panel
dropped to its own idle screen and stayed there, the log filled with send() called before connect(), and only restarting TRCC fixed it, even though the
device was connected and command-line sends worked fine. (#254)
Diagnostic reports keep everything the panel says. A screen introduces
itself with 1024 bytes and TRCC kept 64 of them. Those bytes may well say what
size the screen is — which would mean new coolers being supported by reading
the answer rather than by someone measuring their panel by hand. Nobody could
check, because the bytes were thrown away before anyone could look. They are
kept now.
Documentation fixes: the Gentoo package no longer suggests a trcc detect --all flag that does not exist, and the README's trcc serve example no
longer shows an option the command never had, nor the wrong default port. The
check meant to catch exactly this had stopped reading the files it was
supposed to be checking. (#247)