GitLab

Setting it to 0 would set it to 50 instead of disabling it.

Forces very slow mouse movements to be registered as slightly faster.
This can be useful in games that don't register mouse movements that are
slower than a certain threshold. Flight of the Amazon Queen is an
example of such a game. Do not enable this option in games that don't
have this problem. It makes mouse movement worse.

What this does is simply clamp the values passed to Mouse_CursorMoved()
so that they can't be between -1 and 1. It doesn't fix the buggy games,
but can make things a bit less annoying.

They're just stubs now.

The old name has always been unfortunate, since we don't use nor need
the dosbox mapper.

git-svn-id: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4480 ed541006-0bf7-43e9-8c4d-6fc63c346d47

It makes no sense to call the ARM dynrec cores "generic", as it makes it
sound like there's something wrong. Only do that for x86(64) where the
generic cores are actually sub-optimal and the optimized cores are
preferable.

git-svn-id: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4479 ed541006-0bf7-43e9-8c4d-6fc63c346d47

git-svn-id: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4478 ed541006-0bf7-43e9-8c4d-6fc63c346d47

Because that's what it actually is. The "new" in the name implies it's
better, while actually it's worse (slower.)

kcgen authored Jun 07, 2022 and Nikos Chantziaras committed Jun 07, 2022

Fixes a bug that causes MIDI notes to continue playing when DOSBox
is closed while MIDI is currently playing.

Thanks to @BPaden for reporting this and providing the MIDI message
sequence to properly stop each channel's notes and controllers.

Ref: Mido - MIDI Objects for Python, https://github.com/mido/mido

This is actually the default in upstream. Not sure why we had it
disabled in the core. It has possible perf benefits.

It's a simple hack. We just force a different vtotal and hope for the
best. But it seems to work surprisingly well. Thanks @vanfanel for
thinking of this and suggesting it!

Setting it in retro_init(), even though it works, is actually not
allowed by the spec. The spec says to set it either in retro_load_game()
or retro_get_system_av_info().

It's just too problematic with some retroarch drivers and the benefit
from it isn't that big.

Support BIOS access with multiple images. Fixes some mutli-floppy installers: Dark Seed, Master of Magic, et al.

Implement INT 25h and INT 26h. Fixes some installers: Dark Seed, Gods, et al.

Allow low-level hacks for local drive HDD when floppy image(s) mounted. Fixes installing Amberstar from floppy images.

Use real pointer when writing faked sector contents.

Prevent problematic condition: booting floppy image files when floppy image(s) mounted.

FAT drive improvements regarding zero-length truncation of files.

Update Drive Parameter Table only when necessary.


git-svn-id: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4477 ed541006-0bf7-43e9-8c4d-6fc63c346d47

git-svn-id: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4476 ed541006-0bf7-43e9-8c4d-6fc63c346d47

See: https://github.blog/2022-04-12-git-security-vulnerability-announced

Otherwise build fails with undefined FLAC symbols in MXE builds.

subrepo:
  subdir:   "libretro/deps/fluidsynth-sans-glib"
  merged:   "11ce6d1931"
upstream:
  origin:   "https://github.com/realnc/fluidsynth-sans-glib.git"
  branch:   "v2.2.7-sans-glib"
  commit:   "11ce6d1931"
git-subrepo:
  version:  "0.4.3"
  origin:   "https://github.com/ingydotnet/git-subrepo"
  commit:   "2f68596"

Otherwise cmake might look for libraries in the system path first.

Link with "-Wl,--no-undefined" on Linux and "-undefined error" on macOS
to catch library linking mistakes in the future.

Not sure why this is working fine on some architectures while on others
we get "undefined symbol" when trying to load the core. In any event,
the linking order was wrong and libsndfile not getting linked should
have resulted in errors on all systems. Who knows what was going on.