Bit 7 set for sound in v3 story files

[KVonGit]

When using PunyInform to create v3 story files, if we include calls to @sound_effect, Inform 6.44 will set bit 7 in FLAGS, and the sounds will not play in some older DOS terps, along with Windows Frotz 1.27, which are all checking for bit 4 in FLAGS when loading v3 games.

[erkyrath]

Anyone want to comment on the status of this?

There's been discussion in @DFillmore's https://github.com/iftechfoundation/Z-Machine-Standard repo about @sound_effect, but I don't see anything about FLAGS.

The 1.1 branch of that repo (which still hasn't been declared official) has a line saying:

(flags2 bit 4:) Set in the Amiga version of The Lurking Horror so presumably to do with sound effects?

https://github.com/iftechfoundation/Z-Machine-Standard/blob/d46676270e0562c246b143870693f806d090a4bd/sect11.html#L94

[heasm66]

Additional info for the knowledge base...

Frotz have these definitions for flags in frotz.h

#define SCRIPTING_FLAG	  0x0001 /* Outputting to transcription file  - V1+ */
#define FIXED_FONT_FLAG   0x0002 /* Use fixed width font               - V3+ */
#define REFRESH_FLAG 	  0x0004 /* Refresh the screen                 - V6  */
#define GRAPHICS_FLAG	  0x0008 /* Game wants to use graphics         - V5+ */
#define OLD_SOUND_FLAG	  0x0010 /* Game wants to use sound effects    - V3  */
#define UNDO_FLAG	  0x0010 /* Game wants to use UNDO feature     - V5+ */
#define MOUSE_FLAG	  0x0020 /* Game wants to use a mouse          - V5+ */
#define COLOUR_FLAG	  0x0040 /* Game wants to use colours          - V5+ */
#define SOUND_FLAG	  0x0080 /* Game wants to use sound effects    - V5+ */
#define MENU_FLAG	  0x0100 /* Game wants to use menus            - V6  */

That seem to support that, for v3, bit 4 was used to indicate sound.

Comparing Infocom's own documents.

spec_zip (v3):
 Bit #  Interpretation
 15-2   reserved
 1      fixed-width font needed
 0      script bit

spec_ezip (v4):
 Bit #  Interpretation
 15-3   reserved
 2      request for status line refresh (set by EZIP only)
 1      fixed-width font needed
 0      interpreter currently scripting (set by EZIP only)

spec_xzip (v5):
 bit #   name   interpretation
 0       %FSCRI interpreter currently transcripting
 1       %FFIXE fixed-width font needed (EZIP)
 2       %FSTAT request for status line refresh
 3       %FDISP game uses display operations
 4       %FUNDO game uses UNDO
 5       %FMOUS game uses mouse
 6       %FCOLO game uses colors
 7-15           reserved

spec_yzip (v6):
 bit #   name   interpretation
 0       %FSCRI interpreter currently transcripting
 1       %FFIXE fixed-width font needed (EZIP)
 2       %FSTAT request for status line refresh
 3       %FDISP game uses display operations
 4       %FUNDO game uses UNDO
 5       %FMOUS game uses mouse
 6       %FCOLO game uses colors
 7       %FSOUN game uses sounds
 8       %FMENU game uses menus
 9-15           reserved

An educated guess could be that when Infocom introduces sound in The Lurking Horror (v3) they used an unused flag to indicate sound. This never got codified and this flag later got repurposed for indicating UNDO. In v6 %FSOUN officially got assigned bit 7. I don't think there are any Infocom v5 games that use sound, but I guess that the Z-machine standards backdated this to v5, because it's useful to allow sound in v5 games.

That said, I don't think the compiler should set this flag by default in v3 until it's in the official standards document. The compiler should compile in accordance with the documentation. It could possibly be exposed as an option switch, otherwise the game author could always set this bit manually, in code.

[KVonGit]

Frotz also has this in sound.c (and this is really the only reason I created the issue, because I think this is the bit of code which does not enable the sound in v3 games with bit 7 set in FLAGS):

https://gitlab.com/DavidGriffith/frotz/-/blob/master/src/common/sound.c?ref_type=heads#L73

	if ((z_header.flags & SOUND_FLAG)
	    || (z_header.version == V3 && (z_header.flags & OLD_SOUND_FLAG))

EDIT

I cited the wrong code here. This the code from Frotz, not Windows Frotz, and Windows Frotz is where it looks for bit 4 in v3 games. The code above (of course) checks for the new or old sound flag and works no matter what. If it's a z3 with bit 7 set, this works fine. I'll find this bit in Windows Frotz, and it should be different...

Here it is in Windows Frotz (sorry about that)

https://github.com/DavidKinder/Windows-Frotz/blob/4a6cb898ee923353d12f535d1a54ca898008b130/Generic/frotz.h#L153-L157

#define OLD_SOUND_FLAG	  0x0010 /* Game wants to use sound effects    - V3  */
#define UNDO_FLAG	  0x0010 /* Game wants to use UNDO feature     - V5+ */
#define MOUSE_FLAG	  0x0020 /* Game wants to use a mouse          - V5+ */
#define COLOUR_FLAG	  0x0040 /* Game wants to use colours          - V5+ */
#define SOUND_FLAG	  0x0080 /* Game wants to use sound effects    - V5+ */

  if ((h_version == V3) && (h_flags & OLD_SOUND_FLAG))
  {
    if (!theApp.GotBlorbFile() || !FrotzSound::Init(theWnd))
      h_flags &= ~OLD_SOUND_FLAG;
  }
  else if ((h_version >= V4) && (h_flags & SOUND_FLAG))

---

EDIT 2

...and I think Sherlock is v5.

https://github.com/historicalsource/sherlock/blob/85ab6e5385446cb1ac8cdfe1007cbb372407c76e/gamesound.zil#L7

[erkyrath]

There's an argument that Inform should be able to set any flag bit, if you really want it, for experimentation purposes.

That would be a separate feature. (Perhaps $ZCODE_HEADER_FLAGS_SET, $ZCODE_HEADER_FLAGS_CLEAR, $ZCODE_HEADER_FLAGS2_SET, $ZCODE_HEADER_FLAGS2_CLEAR.)

[KVonGit]

I used unz to check out the headers in the Amiga (and DOS) versions of Lurking Horror and Sherlock. This may (or may not) be helpful: https://gist.github.com/KVonGit/6ed9b46419d88d671e5a19e7c34b2875

Short version:

• on Amiga, bit 4 in FLAGS (aka Flags 2) is set in both versions of Lurking Horror (v3) which had sound (r219 and r221), and it is the only bit that is set.
• Bit 7 is set in Sherlock (v5). (Bit 4 is also set for UNDO in Sherlock, and those are the only two bits that are set.)
• On DOS, there were no releases with sound, and neither bit is ever set. (Lurking Horror doesn't set anything at all in FLAGS (aka Flags 2), and Sherlock sets bit 4 for UNDO and bit 6 for COLOR.)

---

Also, no bits are ever set in MODE (from ZVERSION) (aka Flags 1). That value is always 0. I do not have Atari versions to check. Everything I have is DOS (plus those Amiga releases for the sound functionality); so this concerning bit 5 in the MODE byte of ZVERSION (from the XZIP spec) might have come from those for all I know:

EDIT: This should be the case in v4+. The terp sets the values. I was supposed to know that.

This spec shows the terp should set bit 5 in MODE (flags 1) when sound is available in EZIP and XZIP,

ZVERSION is interpreted as two bytes, VERSION and MODE. All games produced in XZIP
will have a Z-machine version byte of 5; EZIP games will have a version byte of 4; ZIP games
will have a version byte of 3. Combined XZIP/EZIP/ZIP interpreters will need to have this
information, of course. The mode byte contains eight option bits which vary considerably
between EZIP and XZIP.

bit #  name     EZIP interpretation
0      %ESPLIT SPLIT/SCREEN/CLEAR available (0 = no)
1      %EHINV   Inverse video available (0 = no)
2      %EHBLD   Bold available (0 = no)
3      %EHUND   Italic/underline available (0 = no)
4      %ECURS   CURSET/CURGET available (0 = no)
5      %ESOUN   SOUND available (0 = no)
6      reserved
7      reserved

bit #  name     XZIP interpretation
0      %XCOLOR  COLOR operation available (0 = no)
1      %XDISPL  DISPLAY operation available (0 = no)
2      %XBOLD   Bold available (0 = no)
3      %XUNDE   Italic/underline available (0 = no)
4      %XMONO   Monospace style available (0 = no)
5      %XSOUN   SOUND available (0 = no)
6      reserved
7      reserved

For completion, the ZIP spec concerning MODE (Flags 1):

ZVERSION is interpreted as two bytes. Regardless of the state of the byte-swap mode, the
version byte is always first followed by the mode byte. The Z-machine version described in this
document is 3 [SWG 1/13/84], which is the content of the version byte. The mode byte contains
eight option bits as follows:

Bit # Interpretation
7-6   reserved
5     splittability: SWG 1/13/84
4     status line existence
3     Tandy bit
2     machine-specific function
1     status line format
0     byte-swap

---

Also, I'm so new to this, after using a hex editor to change the value from 128 to 16 in the story file a few times during testing, I decided to just modify the inform6 source code to set bit 4 instead of 7, just to use for compiling to v3. I never even considered adding a check for a constant to be defined if I wanted it to set bit 4. That's pretty smart! (I am slowly but surely learning how to actually use my brain, thanks to everyone in the IF community.)

[DavidKinder]

If we're reasonably confident that the use of sounds is indicated by bit 4 in flags 2 in V3, and bit 7 later, at least until there's some evidence otherwise, I think that all that would be needed would be to change asm.c to include two entries in opcodes_table_z[] for sound_effect:

    /* Only two Infocom games used sound: Lurking Horror, a V3 game, used
       bit 4 of flags2 to indicate this, and Sherlock, a V5 game, used bit 7. */
/* 38 */ {(uchar *) "sound_effect",    3, 3, -1, 0x35,      0,      0, 4, VAR },
/* 39 */ {(uchar *) "sound_effect",    4, 0, -1, 0x35,      0,      0, 7, VAR },

We would then have to adjust the comments giving the index of every subsequent opcode. I've not tested this as yet, though.

[heasm66]

One thing to consider is that bit 7 was set in earlier versions of the compiler, and that bit 4 is not mentioned in the z-machine standards (yet...). Theoretically there could be interpreters out there, built according to the standards, that checks bit 7. IF this is done, maybe both bit 4 & 7 should be set for v3.

The line before:

/* 38 */ {(uchar *) "sound_effect",    3, 0, -1, 0x35,      0,      0, 7, VAR },

[erkyrath]

If we're reasonably confident that the use of sounds is indicated by bit 4 in flags 2 in V3, and bit 7 later,

This is the part I'm not very confident about.

I agree that the code change is not a big deal. (It could just as well be a special case where flags2_requirements is set, line 1113.)

[DavidKinder]

#350 adds options to clear and set bits in flags1 and flags2. While this does not directly solve this issue, it at least allows it to be worked around when compiling to Z-code V3.