CTD from Suna Seni (Mac)

[Gorilym]

Yeah, this is a headscratcher. It's the first time I've ever seen the OS X and x86 engine variants differ so wildly on an identical file. If you two are willing to let me experiment, I'd like to see exactly what changes can be made on the OS X side without inducing CTDs--I'd rather pin this down for the next release, and soon.

 

Sure, at this point I'm very curious myself as to what could be causing this.

[Guest s w raly]

Sounds good to me too.

[CamDawg]

After hashing this out over email, it appears we've solved the problem.

 

The confusion immunity effects batch is mis-indexing the effects on items with more than one ability--i.e. Arbane. The OS X variant of the engine is apparently more sensitive to this than the PC one, as it was only causing CTDs on OS X. This is fixed locally and I'm reviewing code to see if we have similar issues elsewhere. Gorilym's Suna didn't have Arbane equipped, which is the real reason it was not causing crashes (rndtre02 was a red herring).

[devSin]

Cast the following on a fixpack install:

COPY_EXISTING_REGEXP GLOB "^.+\.ITM$" "wherever" "^.+\.SPL$" "wherever"
PATCH_IF (SOURCE_SIZE > 0x72) BEGIN
 READ_ASCII 0x0 h (3)
 READ_LONG 0x64 ao
 READ_LONG 0x6a eo
 READ_SHORT 0x70 ix
 FOR (READ_SHORT 0x68 na; na; na -= 0x1) BEGIN
READ_SHORT ao + 0x1e ne
READ_SHORT ao + 0x20 x
PATCH_IF (ix != x) BEGIN
  PATCH_PRINT "%SOURCE_FILE%"
END
PATCH_IF !("%h%" STRING_COMPARE_CASE "ITM") BEGIN
  ao += 0x38
END ELSE ao += 0x28
ix += ne
 END
END
BUT_ONLY_IF_IT_CHANGES

Effect indices must always be correct in the Mac version; I use this code very frequently (part of my permanent devtools TP2).

[CamDawg]

Heh, already wrote my own. Since you've already coded some NI stuff, what about adding a check to the NI tools? I was going to post this request for Fred anyway.

 

The only places this was occurring was on the confusion and hold effects suites for items (propagation via copy-and-paste), and only on ones with multiple abilities. Some of the spells to which we were adding headers had an issue as well. All of these have been fixed locally.

[devSin]

I've been away from it for too long to start up again, so do drop a request to Fred at the forum. My fear is that it will require working with AbstractStruct and co., which is a fat pain in the ass.