Using the right combination of "flags", this addresses the
SQ-INIT error on 10th style as well as forces 10 to 13 into
using the D01 operation mode. The latter follows the approach that
14+ is also using the best/latest IO type supported (IO2 in that
case). Expecting that gameplay experience benefits from more
optimized/better maintained code paths.
Make 10 operate in D01 IO mode which forces security to using
the D01 dongle from a D01 dedicab setup. The alternative would
be using a C02 dongle from a C02 upgraded cabinet.
More details to follow in another commit with a dev journal
entry.
Second part to fixing the 10th style SQ-INIT error. This has
cross-impact on the kind of security data the game expects.
Therefore, we need to expose the IO board type and make it
configurable like the security data for the different game
versions.
More details explained in a dev journal entry to following
in one of the next commits.
This is one part of fixing the 10th style SQ-INIT boot error.
For some reason, 10th style calls the "second dongle slot"
which, according to the code, is expected to return the same data
as the first (black) dongle slot. Original ezusb/C02/D01 IO boards
do not have a second dongle round plug slot nor ever came with
a splitter cable. I checked original manuals of the C02 conversion
kit, E11 and ECO software upgrade kits. No white dongle nor
additional hardware for any other dongle than a black dongle.
The paths changed for 18 and again for 19. Refactor and improve
the detection logic to consider this. This also should make
the code more maintainable.
The synchronization block was too coarse and kept the lock
acquired even when the hooked call continued traversing because
it skipped the chart patch trap code. This was apparently fine
on several games but caused very flaky deadlocking on iidx18
during the boot screens and even crashed iidx19 consistently
before showing the boot screen.
Applies to iidx18 and iidx19 when using the recently extended
redirect feature for settings paths, mixing / and \ crashes the
game in some operator sub-menus when trying to save settings.
The fedora image was already quite old but also took ages to
initially install the dependencies. Being more familiar with debian,
initial container building with package installation being rather
quick, and assuming most folks being more familiar debian as well,
I decided to shift to make this easier to maintain.
- Use static assert to verify struct sizes
- Replace "magic numbers" with proper sizeof's
- Allocate largest size buffer for message response to address
compiler warnings which indicates potential out of bounds
reads/writes
This removes the need to copy all the data into the docker container
and outputs from the build process out of the container. In
hindsight, this was a poor design decision as it created docker
images that were very large with data, that you don't have to
keep stored once the build finished, e.g. the build output.
Therefore, mount the local folder, keep the checkouts also local
which is good for debugging and also re-using them to build without
the docker container (no second checkout required).
Build output is also written directly to the build/ output folder
Overall, the docker container is more lightweight and actually
a "build environment" only not tainted with specific versions
of the dependencies to build.
Provide an option for non final lincle revisions and
if the "modern" monitor check is not delivering stable
results on the final lincle revision, e.g. using the
original "pendual lcd" monitor.
The default should be targeting emulation on a "normal
desktop pc". The default confits for 28 and 29 refeclt
that already, but it seems that 27 was missed out here.
When the process requests to exit, the debugger thread attached
from inject swallowed that event simply without taking action
of actually terminating the process. Handle this accordingly
when the debugger loop is running and terminate the process
