Heretic & Hexen - maps with empty REJECT don't function properly, desync demos

[kitchen-ace]

Here's E1M1 of Heretic with the REJECT lump replaced by a 0-byte lump. Crispy Heretic doesn't behave correctly with this map: if you walk out to the dock in front of you, the gargoyles don't react on seeing you, unless you fire. Demo playback is affected, with demos recorded on empty REJECT maps not even playing back in Crispy Heretic itself, nor DSDA-Doom.

This is the case in both the latest stable version and latest git version, though demos playback differently between the two. (The gargoyles do seem to notice you in stable, though other enemies potentially may not.)

ht-e1m1-noreject.wad.zip

Hexen is similarly affected, Doom is not. I didn't test Strife.

[kitchen-ace]

Here's a 40-second demo on that file using dsda-doom

ht-norej-11-dsda.lmp.zip

[Noseey]

After a first look, it seems that introducing the P_LoadReject from Doom considering the minimum-length check and calling it accordingly, fixes the issue and makes the demo sync again:

P_SetupLevel:

    //rejectmatrix = W_CacheLumpNum(lumpnum + ML_REJECT, PU_LEVEL);
    P_LoadReject (lumpnum+ML_REJECT);
    P_GroupLines();

added P_LoadReject and commented out Padding:

static void P_LoadReject(int lumpnum)
{
    int minlength;
    int lumplen;

    // Calculate the size that the REJECT lump *should* be.

    minlength = (numsectors * numsectors + 7) / 8;

    // If the lump meets the minimum length, it can be loaded directly.
    // Otherwise, we need to allocate a buffer of the correct size
    // and pad it with appropriate data.

    lumplen = W_LumpLength(lumpnum);

    if (lumplen >= minlength)
    {
        rejectmatrix = W_CacheLumpNum(lumpnum, PU_LEVEL);
    }
    else
    {
        rejectmatrix = Z_Malloc(minlength, PU_LEVEL, &rejectmatrix);
        W_ReadLump(lumpnum, rejectmatrix);

        //PadRejectArray(rejectmatrix + lumplen, minlength - lumplen);
    }
}

[mikeday0]

I'm guessing this probably ought to be fixed upstream in Chocolate.

[fabiangreffrath]

I'm guessing this probably ought to be fixed upstream in Chocolate.

Yes, right. Two things: (1) P_LoadReject() should be moved from Doom to a more local place so that all four games can make use of it and (2) this should happen upstream in Chcolate Doom.

[Noseey]

I will look into it and create a PR for Chocolate Doom.

[fabiangreffrath]

I will look into it and create a PR for Chocolate Doom.

Thank you! I think you can just factor out the PadRejectArray() function into something like src/p_rejectpad.c and you'll only have to additionally pass totallines to it.

[Noseey]

Hm, seems like there is more to it. I tried the exact same hotfix mentioned above in chocolate, but it didn't resolve the desync there. I will investigate further..

[kitchen-ace]

Is adding this kind of fix to Chocolate not a deviation from Vanilla behaviour anyhow?

[fabiangreffrath]

Hm, seems like there is more to it. I tried the exact same hotfix mentioned above in chocolate, but it didn't resolve the desync there. I will investigate further..

Crispy might handle intercepts in Heretic differently than Choco.

Is adding this kind of fix to Chocolate not a deviation from Vanilla behaviour anyhow?

It adds emulation of the engine reading bytes past the memory allocated by the REJECT lump. Choco already does this for Doom and Strife, this is about adding the same to Heretic.

[mikeday0]

Hm, seems like there is more to it. I tried the exact same hotfix mentioned above in chocolate, but it didn't resolve the desync there. I will investigate further..

Don’t you still need to fill the reject array with zeros? That’s part of what PadRejectArray does but I noticed it was commented out above.

[fabiangreffrath]

Don’t you still need to fill the reject array with zeros? That’s part of what PadRejectArray does but I noticed it was commented out above.

That function is static in Doom and Strife, and I guess that's why @Noseey has commented it out in the example above. That's also why I suggested to factor it out and make it available to all four games.

[Noseey]

Sorry for the confusion, it was a very silly issue with my setup that I'd rather not want to tell... I can confirm the hotfix also works for chocolate, both with or without padding. I will proceed with the PR tomorrow..