Some games repeatedly query for the first track's position during
playback, perhaps as a dual-purpose "CDROM health/heart-beat"
check.
This excessive console output in DEBUG mode tends to add far more
noise than signal, and prevents the maintainer from seeing the
overall flow of the CDROM calls.
Moves to using a member vector to provide a persistent buffer,
instead of repeatedly allocating and deleting memory on every
invocation (during DAE, this can be called hundreds of times
per second, with requsts of up 24 sectors or ~56KB of memory
per call).
The buffer is only resized upward, which avoids 'zig-zag'
re-allocations when switching between raw (2352-byte) and cooked
(2048-byte) sectors.
Only writes as much data was was successfully read into
the DOS program's buffer, avoiding (potential) garbage-fill.
This commit:
- Adds more sanity checks and comments
- Lets the seek() function to take care of monitoring the track's
post-seek decode position, instead of managing it in read(..).
- Adds in-place conversion from mono-to-stereo
- Adds micro-second-level DEBUG timing around the mono converter
- Makes use of the more intuitive ceil_divide() support function
to avoid excessive casting and floating-point conversion.
This commit:
- Comments why we convert from byte-to-time offset.
- Checks for underlying stream validity before seeking.
- Tracks the new as-seeked decode position.
- Skips seeking if it's unecessary.
- Adds DEBUG messaging and a warning if the seek took
longer than that of an average physical CDROM.
The commit refines several types to their logical use-cases.
For example, a CDROM will at-most contain 400,000 sectors and be
less than 1B bytes in size. Likewise, both are 'physical'
quantities and this should always be zero-or-greater
(therefore, uint32_t is used).
Fortunately, there are almost no cases where these values are
overloaded to mean something else (ie: negative return codes to
indicate failure).
Besides eliminating many implicit cast warnings, differing-signed
comparison warnings, and overflowable type-cast warnings, the
more correct use of types helps logically bound expectations of the
values they contain, which should improve maintainability.
1. Moves the mutex and mixer channel members to unique pointers
2. Moves the trackFile to a weak pointer
3. Move member initialization to the class definition
This class still retains a raw *cd member, however fixing this
ripples up to the array of [26] CD images, and across more
code that generically deals with mount points; so this work
remains to do.
Most MinGW versions work correctly with and without _POSIX_C_SOURCE
or XOPEN macros [1] and don't need any special handling, but some
installations refuse to work correctly.
To be more precise: *some* MinGW installations turn off POSIX support
after turning on -std=c++11 WITHOUT -ansi -Wpedantic, -pedantic-errors
and similar. This behaviour is inconsistent between various Windows
versions and MinGW distributions and manifests itself only when using
GCC (Clang behaves correctly).
[1] https://www.gnu.org/software/libc/manual/html_node/Feature-Test-Macros.html
Many DOS games open all their files in write-mode regardless if the
game actually intends to write to them or not.
This can result the files' metadata being updated (lossing the
original date-stamp) as well as putting the file at-risk of
corruption if the game crashes with the file open for writing.
Under the existing DOSBox implementation, if a user attempts to
write-protect their data files then DOSBox will quit with an error in
the above scenario.
This commit allows the use of write-protected files (similar to
running a game from CDROM) and indicates which files are being
actively protected against modification.
Typically the only files that need to be left writable are those that
the game actually changes, such as: save-game files, highscore
files, in-game settings, and so on. If these are also protected,
then the game will quit/crash immediately after the protected message
is printed, and thus indicate which file require write-allowance.
This code made silent assumption that first fields in direntry are
exactly 14 bytes - this was fine, except would break as soon as anyone
would touch the struct (or e.g. if a compiler would lack support for
packed structures and inject some padding in there); rewrite the copy
code to follow the same pattern as other fields - now the code will be
fine even if someone will change fields in the direntry struct.
Fixes 2 PVS static analysis issues (buffer underflow on src and dst).
Effect is the same, except packed values are not referenced via pointer
so there's no address-of-packed-member warning, and no need for
unaligned memory access.
This function uses its boolean return code to indicate success
or failure, which all callers check prior to using any of the
referenced arguments. Therefore it's unecessary to zero-out
referenced arguments or use intermediate variables.
The specification says that legal track values range from 1 to 99,
where as the prior code would return 0 if any issue was encountered.
The spec has no allowance for issues in this function, and therefore
we're bound to simply return 1 instead.
For example, audio CD players will typically position themselves in
preparation to play the first audio track when "Play" is pushed.
This even occurs for mixed-mode CDs, where the player will start
playing at the first audio track. Therefore, we try to find the most
viable audio track and only fall back to defaults if needed.
(instead of "all zeros" like before).
With types refined, we can switch to cleaner integer
math and avoid casting (to float and back to ints), and also
avoid ceil((float) a / b) by using the pure-integer form
of (a + b - 1) / b
This is simplified by no longer retaining the read position,
which is unecessary because all read operations are offset
from an initial absolute position (so this was unecessary code).
Bitu (aka uintptr_t) is not necessary for handling shortened files,
as we can represent max 10^7 shortened files anyway; unsigned is good
enough.
This allows use to use simple %u for conversion instead of PRIuPTR,
which caused a bug on AmigaOS on PPC.
Fixes: #162
Class DOS_Drive_Cache is declared in dos_system.h, but this header was
indirectly included. Remove faulty windows.h include (the correct one
exists in cross.h already).
