Measured rather than guessed: the whole repository was cross-compiled for
linux/amd64 and the gaps closed one by one. There were fewer than expected.
Flex and TCI were never Windows-specific — they carried //go:build windows by
inheritance and import nothing but net and gorilla/websocket. Untagged, no code
change. The two backends a Linux operator is most likely to own were already
portable.
Audio was 560 lines, not 2287: only devices.go and engine.go touch WASAPI, while
manager.go, recorder.go, wav.go and mp3.go were pure Go wearing the tag by
association. The whole platform surface is seven functions, now implemented a
second time on PulseAudio through github.com/jfreymuth/pulse — pure Go over the
server socket, so the no-cgo rule survives, and PipeWire answers the same
protocol. The fixed 16 kHz mono format and the server-side resampling mirror
what AUTOCONVERTPCM does on Windows, for the same reason.
OmniRig is the only real loss, and its backend still EXISTS off Windows rather
than being compiled out of app.go: a settings database is portable, so an
operator moving a profile across keeps "omnirig" saved and must be told to pick
a native backend instead of meeting a nil one.
The parts where Linux is not Windows, and where a compile-only stub would have
been a silent bug:
- data dir: still beside the binary, but ~/.local/share/OpsLog/data when that
folder belongs to the system — decided by trying the write, because /opt and
/usr/local are writable on some stations and not others.
- single instance: an flock, not a pid file. The kernel drops it however the
process dies, so a crash leaves nothing to delete by hand. This is the guard
that stops two instances fighting over the rig frequency.
- update: simpler here. Unix renames over a running binary, so the deferred
swap the Windows path needs a detached helper for is unreachable.
- tasklist/taskkill become /proc and SIGTERM; the boot log moves out of /tmp,
which is wiped exactly when the evidence is wanted.
- serial ports sorted naturally: /dev/ttyUSB10 was landing between USB1 and
USB2, the same trap COM10 fell into.
release.ps1 now cross-builds and vets for linux before it builds the exe, and
refuses the release if that fails — a port rots one unguarded x/sys/windows call
at a time.
Nothing has been executed on Linux yet: Wails needs webkit2gtk and cgo there, so
the binary must be built on Linux. scripts/linux-setup.sh checks the machine and
does it; BUILDING-LINUX.md is the manual version.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The relaunch after an update stopped working, and the regression is mine:
removing the PowerShell helper — which is what Defender was reading as a
dropper — also removed the wait it was doing. Nothing took over the job.
The numbers made it certain rather than unlucky. The instance being
replaced is allowed THIRTY seconds to shut down (armExitWatchdog forces
it out at that point) because it closes a remote logbook, a CAT session
and sometimes a backup. The new instance was patient with the
single-instance mutex for TWENTY. On any station where shutting down ran
past that, the new process gave up and exited in silence: no window after
an update, and the previous OpsLog still in the task manager. Exactly the
report.
Both relaunch paths now pass --wait-pid, and the new process waits on
that process's handle — a plain kernel wait, which ends the instant the
old one ends, however long or short that is, and looks nothing like a
script starting another program. The mutex retry stays as a backstop and
goes to forty-five seconds, so it is longer than the wait it exists for
rather than shorter.
And when the old process really has not gone, the message says that
instead of "OpsLog is already running" — after an update the operator did
not start a second copy, and what they need to know is which one to
close.
A test keeps the two spawn sites honest: a relaunch added without
--wait-pid is this bug again.
rouggyandClaude Opus 5