I forgot to test the pyinstaller binary when I upgraded all the other
dependencies, and there's a breaking change in 6.0.0:
https://pyinstaller.org/en/latest/CHANGES.html#incompatible-changes
> All of onedir build's contents except for the executable are now moved
> into a sub-directory (called _internal by default). sys._MEIPASS is
> adjusted to point to this _internal directory. The breaking
> implications for this are:
>
> * Assumptions that os.path.dirname(sys.executable) == sys._MEIPASS
> will break. Code locating application resources using
> os.path.dirname(sys.executable) should be adjusted to use __file__
> or sys._MEIPASS and any code locating the original executable using
> sys._MEIPASS should use sys.executable directly.
> * Any custom post processing steps (either in the .spec file or
> externally) which modify the bundle will likely need adjusting to
> accommodate the new directory.
This is actually great because it declutters the top level directory to
just `dcs_liberation.exe` and a directory named `_internal` that has all
the guts, but the CWD is no longer the directory that has `resources/`
in it, so we can't find any of our resources. There are a few options
for fixing that (cd into that directory probably being the easiest, or
we could stop relying on CWD relative paths), but for now just downgrade
to unbreak the build.
A lot of the dependency versions we have pinned don't have wheels for
Python 3.12. Update almost all of them so we can upgrade Python.
The few that weren't upgraded here are black and mypy, since those will
be a bit invasive, and Pillow, which has an API change I don't want to
deal with right now (I've got a commit on another machine that has
already done the migration, so I'll do it later).
Now that the F-15E Suite 4+ has JDAMs and the bug preventing AI from
using LGBs has been fixed, I have removed the AI-only F-15Es from my
campaigns. Also minor tweaks to a couple of squadron types/sizes based
on play-testing results.
There are a few different code paths for deleting packages depending on
the state of the package, and two of them were deleting items from a
list they were iterating over without first making a copy, causing each
iteration of the loop to skip over a flight, making it still used since
the flight was never deleted, but unreachable since the package that
owned it was deleted.
This only happened when the package was canceled in its draft state
(the user clicked the X without ever saving the package), or if the user
canceled a package mid fast forward (the controls for which aren't even
visible to users) while only a portion of the package was active. In
both cases, only even numbered flights were lost.
There's a better fix lurking in here somewhere but the interaction with
the Qt model complicates it. Fortunately that mess would be cleaned up
automatically by moving this dialog into React, which we'll do some day.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/3257.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/3258.
It appears the AI is still incapable of using LGBs (and laser JDAMs),
even though Razbam had said the issues was fixed. I have switched
loadouts to JDAMs because the AI will use those.
This fixes a map object strike target (yes, just one. Fortunately!) that
was broken by the latest DCS open beta update that removed the buildings
the strike target was using.
GBUs have been added to the F-15E Suite 4+ loadouts
Also switched CAP loadouts to 2 external tanks instead of 3 to improve
aircraft performance, as 2 tanks is plenty for players and the AI has
infinite fuel now.
This adds LARC-V to Liberation (required by USA 1970)
Also adds Speedboat to Liberation (required by NVA 1970)
Both units are available via pydcs but have not been added to Liberation
thus far as no factions used them (until now).
This is enabled by default because I can't think of a time where it's
ever been more fun to watch the AI run out of fuel after cycling between
afterburner and speedbrake for twenty minutes.
This is only lightly tested (I verified that the task shows up
appropriately in the ME when set, not when unset, and never for
players), since the interesting part of the implementation is up to ED.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/3201.
Updates the S3 Viking so they can now do DEAD (since the new model can carry SLAMs). Also updates the loadouts for all the other mission types as they are outdated (2000 pounders for anti-runway, 500 pounders for strike, Mavericks for BAI/CAS, etc).
We had two different paths for converting pydcs loadouts because pydcs's
APIs for some reason return the loadouts in different shapes which made
it difficult to share the code for converting them. Rather than fix the
bug in both places, extract the common code and adapt the result of one
API to match the other.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/3171.
There's more detail in the comment, but this fixes an issue where some
Mosquito loadouts could not be loaded.
Might fix https://github.com/dcs-liberation/dcs_liberation/issues/3171.
There aren't enough instructions in the bug for me to be sure, but it
sounds like a similar problem, although I came across it with bombs
rather than rockets.
