Added support for InstallShield v6 cabinets, which can be made up of
multiple files. The interface for creating an Archive instance now takes
the base filename (e.g. the "data" in "data1.cab") for all cabinets,
including single-file ones.
Co-Authored-By: clone2727 <236052+clone2727@users.noreply.github.com>
Co-Authored-By: Walter van Niftrik <615114+waltervn@users.noreply.github.com>
OSystem now just returns a nullptr if there is no text to speech manager instance
(because none is compiled into the binary, or the system doesn't provide support
for it). This removed the need for the engine authors to add scummvm osystem compile
time options checks into their engine code
CJK offsets were wrong compared to encodings.dat file and buffers were
read past their end.
Rewrote indexing to make it match the Python script and optimize
slightly
We have a lot of legacy code that was written with the old behaviour
where it returned the default value in such a case. Until we are
confident all this code has been updated, we continue to use the
old behaviour in release builds to avoid creating instabilities.
This still error our in non-release builds, which will help detect
the code that still needs to be updated.
The string memory pool mutex lock/unlock were lost during the
merge of String and U32String. This caused the cloud feature
to randomly crash for example when synchronizing savegames or
downloading game data.
With __printf__, MinGW prefers the portable old ms-style printf, and
warns if formats like %lld and %hhd are used.
In practice, if __USE_MINGW_ANSI_STDIO is not defined, it defaults to 1
on C++11, and these formats are supported. Using __gnu_printf__
suppresses bad format warnings.
`convertBiDiStringByLines` calls the BiDi algo for each line in isolation,
and returns a joined result.
That's needed to support BiDi in AGI, and might be needed for other engines
in the future.
In order to do that, a new utility function was added:
`String::forEachLine` which gets a function as input, and its arg(s) (if it has any),
and calls the function on each line, and returns a new string which is all
concatenation of all the lines results (with '\n' added between them).
This is emitted when -Wdeprecated-copy is enabled. The fix implemented
is to explicitly declare this to be the default copy operator.
Since this may be a latent bug, this is marked with a FIXME.
bits/os_defines.h has #define NOMINMAX 1, which conflicts with the
non-value #define NOMINXMAX in scummsys.h.
In file included from ../scummvm/common/lua/lua.h:12,
from ../scummvm/common/lua/ltable.cpp:27:
../scummvm/common/scummsys.h:111: warning: "NOMINMAX" redefined
111 | #define NOMINMAX
|
In file included from C:/msys64/mingw64/include/c++/10.2.0/x86_64-w64-
mingw32/bits/c++config.h:518,
from C:/msys64/mingw64/include/c++/10.2.0/cmath:41,
from C:/msys64/mingw64/include/c++/10.2.0/math.h:36,
from ../scummvm/common/lua/ltable.cpp:21:
C:/msys64/mingw64/include/c++/10.2.0/x86_64-w64-mingw32/bits/
os_defines.h:45: note: this is the location of the previous definition
45 | #define NOMINMAX 1
|
Amends commit 88f685217b.
This config was handled in the DefaultEventManager, but this did not
work for engines that quit the game without using a QUIT_EVENT (such
as SCI). So now it is handled one level up directly in scummvm_main.
Note: It is still also handled in the DefaultEventManager so that we
get the correct confirmation dialog there when "confirm_exit" is true.
Different platforms have different levels of support of encodings and
often have slight variations. We already have tables for most encoding
with only CJK missing. Full transcoding inclusion allows us to get reliable
encoding results independently of platform. The biggest con is the need for
external tables encoding.dat.
It removes a duplicate table for korean in graphics/korfont.cpp
The commit does not change the behaviour of the class but makes
its implementation more logical, and in line with how it was
originaly implemented (before being refactored in commit 1011508).
Both FROM_BE_* and TO_BE_* do the same thingi for example, but you
use this stream class to read from a BE or LE stream and convert
the values to native endianness, and thus it makes more sense to
use FROM_BE_* and FROM_LE_* instead of TO_BE_* and TO_LE_* (which
suggested that we read values in native endianness and then
convert to the specified endianness).
This function used to be called from the Engine class, but was
not longer used since commit 432fd522. The single remaining use
was an internal use in the SurfaceSdlGraphicsManager, and I kept
the behaviour there while removing the function.
Remove a useless creation of a BaseString object with a C string in
comparison operators of this class, because there is a function
equals() that accepts a C string in parameter.
The call to decRefCount has been moved to the destructor
of the base class BaseString.
Note that BaseString only exists for the purpose of being able
to reuse code for our String and U32String implementations. As
such it is not meant to be used polymorphically and its
destructor does not need to be virtual (with the overhead that
this generates). And to ensure that it is not used in a
polymorphic way, the destructor is declared protected.
U32String can contain characters from all languages, without relation
to the GUI's language. Therefore, we need to always call it.
Use case:
In Hebrew SQ3, when using the game's original save/restore dialogs, it's
possible to give save games Hebrew names.
What happens when we then use ScummVM's load dialog?
- If ScummVM's GUI language is Hebrew, it's calling BiDI algo, and
the text direction is OK.
- However, if ScummVM's GUI language is English, before this commit it
didn't call the BiDI algo, and the text was reversed.
Appending \0 to string and expecting it to be just dropped is
still an invalid behaviour but it already happened in 2 engines, so
restore old behaviour, at least for now
Adding @defgroup and @ingroup doxygen tags into all headers
in the common folder that contain doxygen blocks.
This improves the structure, readability, and findability
of information in the resulting output.
This commit targets purely structure and does not deal with
the content of the currently existing doxygen documentation.
The c++11 standard includes some features that do not only depend on the
compiler supporting it, but also the c++ standard library having support
for it. This is the case for the initializer list feature. So when we
compile with a modern compiler but using an old std lib we need a
replacement for it.
Enabled system API for begin/end gfx transactions, initSize, setGraphicsMode.
Function setGraphicsMode will use optional params to trigger 3d rendering
and switch to proper SDL Gfx manager.
Common::MacResMan is now able to open files from a specified
Common::Archive. This is a bit hacky as dynamic_cast is used to break
the Archive encapsulation to retreive the underlying FSNode. It should
however be more correct than the previous code that assumed files were
at the root of the currently running game's path.
AdvancedDetector constructs a Common::Archive from its FileMap based
filesystem cache and uses it to detect the mac resource fork files.
This cuts the time it takes to run the detection code with all the
engines enabled as dynamic plugins on the 3DS to 30 s from 280 s.
They were supposed to compare two U32String, but due to a copy/paste
error they were comparing a U32String with a String. It compiled
because there is a String constructor that takes a U32String (and
converts it to UTF-8), but it was probably not quite working as
expected for non-ASCII characters.
When assigning a String or char * to an existing U32String, the code
assumed that the current _str pointed to the internal storage. But
that is not the case if the U32String contains a string longer
than the internal storage capacity prior to the assignment, and
this led to various memory issues (and usually a crash down the
line).
- Remove = operator in String, which compared to a U32String.
- Let implicit cast take care of results from getResultString, so in-future easy to spot new changes.
- Use uint32 for each char when encodingUrlStrings, to avoid unsafe comparision.
It was not always setting to 0 all the newly allocated bytes,
and if you were unlucky this could result in additional spurious
characters at the end of the converted string.
- Allow the base class to take in a string with custom charset, convert to U32 and pass it along.
- Remove redudandant functions in all tts-subsystems
- Remove unnecessary "charset" argument, as U32Strings imply that we are using a UTF-32 charset
- Adjust relative code appropriately according to the above point.
- Add a contains utility function to ustr
- setFilter uses U32String and U32Tokenizers
- Make consequent changes in launcher to allow sending u32strings
- Common: add wordWrap function to ustr.cpp
- Bladerunner: Explicitly state we have a U32String in subs (same as Subtitles::loadOuttakeSubsText)
- Don't use translations for engine specific "put strings", because they might not support.
- SCI: Use const references for showScummVMDialog
- SCUMM:
-- Don't use translation in md5 warning. left comments with the translated version.
-- Remove some redundant headers in help.cpp
-- Don't use translation in handleSaveload when printing to console
-- Also, display success transaction correctly via u32::format
- TESTBED: Use fake constructor when setting label of button
- SKY: Correctly use translation when using SaveStateDescription
- ULTIMA: Don't use translations when display_string
- ENGINES:
-- GenerateUnknownGameReport correctly, with proper translations.
-- There was an error, where a function had been declared twice, in a header file. Correct this.
This commit addresses a range of changes, within scummvm subproject.
- Audio files, like mididrv, remove U32String based name and identifier, because ASCII only.
- mididrv.cpp had some wrong format for warning messages, fix those
- Message dialogs were modified to use default arguments more often, but reverting back to the orignal to minimize changes.
- SetTooltip has a fake constructor that takes in a string, and use it.
- U32Format had some break statements missing, add those.
- RemapWidget: Use fake constructor for setLabel and setTooltip, to make minimal changes
- SDL: setting text in clipboard no longer uses SDL_iconv_string
- TTS: Override base class "say" with strings, so tts->say can be used with normal strings too.
- About dialog: fix incorrect code for u32string variables
- Fix some extra brackets
- Some buttons were incorrectly removed from using translated labels, revert those
- Message Dialog: Pass default and alt buttons as const references
- Saveload Dialog: Use translations in missing places, use const-references. Also, use translations in a correct manner.
- Use const references for tooltip in GraphicsWidget, EditTextWidget, error.cpp
- DomainEditTextWidget: Use U32String for text
Up until last commit, everything was working fine but the amount of files changed was too large. This commit tries to reduce the changes.
- Add a fake constructor to Keymap, text-to-speech, setDescription (save-state)
- Redirecting functions for PopUpWidget::appendEntry, ButtonWidget::setLabel, GUIErrorMessage
- Use the above functions and constructors to reduce changes in Engines
- Fix warnings being in unicode. Only output english text in - Warnings, Errors, etc.
- Mark some strings as "translation" strings. (Not yet added to POTFILES)
- Remove some CP related things from po/modules.mk
- Previously used some Common::convertToU32 where it was not necessary, replace this with u32constructor
- Revert accidentally put translations in mt32.cpp
- Use U32::format in some places earlier missed
- Add %u and %i for u32::format
- Add support for GUIErrorMsgFormat to use u32::format internally
- Use the above whereever needed
- Improve linux tts by removing redundant code
- Some places I had changed nullptr -> "". Revert this
Because po files now (should) use UTF-8, a bunch of code in ScummVM is redundant. This commit addresses that.
- Remove charsetMapping and related variables.
- Remove the code that used these functions.
- Remove cp_parser.cpp. This is removed because the use of codepages is no longer needed.
- Remove iso-8859-x codepages (see above message)
- Always set mapping as 0 in ThemeEngine::loadScalableFont
- Check if ascii or not by getting charset in about dialog.
- Changes made in create_translations tool and translations.cpp
- After converting po files to UTF-8, length of block did not fit in 16 bits. This commit addresses that.
- The previous commit removes all codepage additions to the create_translations tool.
- Following that, this commit removes all the reading from .dat file related to codepages
- Use the fake constructor for message dialog less often
- solve errors regarding comparision between string and u32string in midi
- solve errors in testbed/speech related to ttsMan->say
- fix unsigned comparision warning in tab.cpp
- fix issue with static strings and translations in sdl-graphics
They always take in U32 strings now.
- Revert tts descriptions to string, they don't use translations, so better to keep it as strings.
- Make read() take in const U32 references.
Only for strings without an explicit action. The first level takes in a u32 string, encodes it, and passes it along further with the default action
- For gui widgets, the strings are passed in as native u32 strings.
- For everything else, they are being converted to u32 by Common::convertToU32... and then being passed along
This commit also includes some additional major changes.
- Return U32String from TransMan getTranslation()
- Change tooltips to be U32Strings
- Remove temporary code of convertToU32String
- U32 Support various dialogs (Browser, FileBrowser, Messages, Chooser, etc)
- U32 for music plugins
- U32 for OSD messages
- Add some helper functions to ustr.cpp
- Change default tooltips from nullptr & 0 -> ""
- Some runtime exceptions may occur when changing languages due to incorrect String::Format
- Previously, if using an RTL-based language and disabled fribidi, console would repeatedly have the same messages spamming.
- Reduce this to only print warning one time.
