Difference between revisions of "Gob/TODO"

From ScummVM :: Wiki
< Gob
Jump to navigation Jump to search
(Two bugs are fixed)
(Update several issues, describe new bugs)
Line 11: Line 11:
* All structs directly mapped to memory that are not trivially replaced (such as 'scen_animations'). Currently use (or should use - this is still WIP) READ_LE_* functions to read most member variables which are known to be in LE format and only read from - NOT written to. Later, when the data is hopefully converted to native format during loading, these READ_LE_* calls should obviously go.
* All structs directly mapped to memory that are not trivially replaced (such as 'scen_animations'). Currently use (or should use - this is still WIP) READ_LE_* functions to read most member variables which are known to be in LE format and only read from - NOT written to. Later, when the data is hopefully converted to native format during loading, these READ_LE_* calls should obviously go.


== Showstoppers/Overall bugs ==
== Bugs ==
* Currently only VGA floppy is supported so we should add MD5-based detection before it will go to CVS
* It hangs on "please wait" screen in EGA version when screen is completed or switched by code
* Keyboard input is buggy as some keypressed get ignored. Particularly noticeable when entering screen code
* When game calls delay() screen doesn't get redrawn. Noticeable at "please wait" screen and especially cutscenes (try screen #5 FTWKFEN)
* Demo doesn't handle last screen properly. It hangs instead of printing ordering info
* Demo doesn't handle last screen properly. It hangs instead of printing ordering info
== More on Endianness and Alignment ==
'''Assignee:''' Carthag.
''I am going to work on this, this Sunday. I will initlally confine myself to game.cpp, and then start migrating the changes to wherever the files are accessed.''
It crashes now in game.cpp function game_loadExtTable() at
data_readData().
This is all bad. Whole game tries to read structures directly without parsing them.
Problems which it arises:
* sizeof(struct) is bigger than sum of sizes of its elements on most systems due to alignment. This is why it crashes now
* Endianness is not honoured
What we should do here is get rid of both data_openData and data_readData()
and write function with this prototype (give it proper name):
  MemoryReadStream data_getData(const char *filename, const char *chunkname);
It will read whole named chunk and return a stream of it. So code using it
will look as follow:
  chunk = new data_getData("intro.stk", "intro.tot");
  foo.a = chunk.readUint32LE();
  chunk.read(foo.b, 20);
  delete chunk;
MemoryReadStream will track end of chunk by itself, so no additional code is needed.
All places should be revisited and rewritten as every one will cause problems.
This will let us get rid of file_open() too which is an implementation of
an ugly idea.
==== Update ====
''sev:'' As a temporary solution I fixed that with such code:
  data_readData(game_extHandle, (char *)&game_extTable->items[i].offset, 4);
  game_extTable->items[i].offset = FROM_LE_32(game_extTable->items[i].offset);
  data_readData(game_extHandle, (char *)&game_extTable->items[i].size, 2);
  game_extTable->items[i].size = FROM_LE_16(game_extTable->items[i].size);
which is ugly, difficult to read but feasible until proper object-oriented implementation will be done


== Main and long-term goals ==
== Main and long-term goals ==
Line 65: Line 24:
=== gob.cpp ===
=== gob.cpp ===
* Write proper game detection. (MD5-based?)
* Write proper game detection. (MD5-based?)
=== interp.cpp ===
* Since neither VGA nor CD versions have protection screen we may skip it in EGA version too


=== map.cpp ===
=== map.cpp ===
Line 71: Line 33:
=== sound.cpp ===
=== sound.cpp ===
* Has yet to be written from sound.asm
* Has yet to be written from sound.asm
* Support Red Book audio in CD versions


=== text.cpp ===
=== text.cpp ===
Line 80: Line 43:
=== video.cpp ===
=== video.cpp ===
'''Assignee:''' JoostP is working on these
'''Assignee:''' JoostP is working on these
* Implement <strike>pDrawSprite</strike>, <strike>pFillRect</strike>, pDrawLine, <strike>pPutPixel</strike>, <strike>pDrawletter</strike>, pDrawPackedSprite functions from the VGA driver (lvga.gdr)
* Implement <strike>pDrawSprite</strike>, <strike>pFillRect</strike>, pDrawLine, <strike>pPutPixel</strike>, <strike>pDrawletter</strike>, <strike>pDrawPackedSprite</strike> functions from the VGA driver (lvga.gdr)
* Classify
* Add support for other rendering modes (EGA/CGA/Hercules) lega.gdr, l360.gdr, lcga.gdr
* Add support for other rendering modes (EGA/CGA/Hercules) - the EGA version of Gobliiins works with gobreverse, so vid_spriteUncompressor needs to be adjusted and/or the packedSprite drawing routine from the driver need to be implemented.

Revision as of 22:11, 11 April 2005

Status

Game is completable. There are occasonal minor graphical glitches. The code is not endian-safe, JoostP is looking into this as time permits.

Problematic places

  • map_loadDataFromAvo() in map.cpp. These are plain calls to memcpy() so all read data is not endian safe and moreover it could read wrong size if sizeof(structure) is used in length calcualtion. All structures should be unwind, i,e, read element by element and FROM_LE_XX wraps applied.
  • data_readData(). Same as above.
  • int32 *gob_curGobStateVarPtr and all gob_*Ptr variables. They all map directly into data, and have to be fixed.
  • All structs directly mapped to memory that are not trivially replaced (such as 'scen_animations'). Currently use (or should use - this is still WIP) READ_LE_* functions to read most member variables which are known to be in LE format and only read from - NOT written to. Later, when the data is hopefully converted to native format during loading, these READ_LE_* calls should obviously go.

Bugs

  • It hangs on "please wait" screen in EGA version when screen is completed or switched by code
  • Keyboard input is buggy as some keypressed get ignored. Particularly noticeable when entering screen code
  • When game calls delay() screen doesn't get redrawn. Noticeable at "please wait" screen and especially cutscenes (try screen #5 FTWKFEN)
  • Demo doesn't handle last screen properly. It hangs instead of printing ordering info

Main and long-term goals

  • Convert it to C++ classes. This should be done on per-file basis and keep in mind that we should make it to resemble original code al little as possible.
  • Endianness fixes

File-specific tasks

gob.cpp

  • Write proper game detection. (MD5-based?)

interp.cpp

  • Since neither VGA nor CD versions have protection screen we may skip it in EGA version too

map.cpp

  • See FIXME comment. All calls to map_loadDataFromAvo() should be made endianness and alignment-safe.

sound.cpp

  • Has yet to be written from sound.asm
  • Support Red Book audio in CD versions

text.cpp

  • This file should be removed and is not currently required for compilation but we need to inspect which variables does it set and to which values and convert them to our Config Manager. We should support all video outputs and at least PC speaker and Adlib.

util.cpp

  • Get rid of rand() and use RandomSource

video.cpp

Assignee: JoostP is working on these

  • Implement pDrawSprite, pFillRect, pDrawLine, pPutPixel, pDrawletter, pDrawPackedSprite functions from the VGA driver (lvga.gdr)
  • Add support for other rendering modes (EGA/CGA/Hercules) lega.gdr, l360.gdr, lcga.gdr