Game crashing on quadro cards

[nobby]

I'm getting crashing symptoms similar to those described in viewtopic.php?f=6&t=3227 on a machine with a Quadro FX1400 video card. The drivers have been upgraded to current as of a few weeks ago and the version is 2.2.1. Hardware is a HP XW9300 (this has an OEM Tyan S2895 motherboard) with Windows XP SP2 + current updates.

(i) Has WZ2100 been tested significantly with Quadro cards - is this a known problem?

(ii) If the answer to (i) is 'No', I have some mid-high end Quadro cards[1] (not the latest but not that old either) that are surplus to requirements. Would the devs be interested in one of these as a donation?

[1] Specifically, a Quadro FX3450, a Quadro FX1400 and (if someone has an older machine with an AGP slot) an AGP Quadro FX3000.

Nigel.

[cybersphinx]

(i) Has WZ2100 been tested significantly with Quadro cards - is this a known problem?

I don't remember any specific reports with Quadros, and in the hardware thread you're the only one who listed one. So I guess there was no significant testing. Have you tried different driver versions, or just your current one? Is there a crashlog (warzone2100.rpt I think), or stdout.txt/stderr.txt files?

(ii) If the answer to (i) is 'No', I have some mid-high end Quadro cards[1] (not the latest but not that old either) that are surplus to requirements. Would the devs be interested in one of these as a donation?

Oh, wow, I think that's the first case of "It doesn't work, can I send you my hardware to fix it?"

Hm, I'd be mildly interested in a card... let's see who else speaks up.

[nobby]

Hi, I've just uploaded a Warzone2100.rpt file and attached it to the original posting - can the forum app support multiple attachments?

I've tried it on two driver versions - whatever was installed on my machine at the time and the latest one as of a few weeks ago. I also played 2.1 (the version that came with Ubuntu 9.04) and experienced similar crashes on that with two versions of the nvidia drivers. That machine was a HP DC7100 with a Quadro NVS285.

[Zarel]

Yeah, the limit is 3 attachments per post.

Crash dumps are pretty useless for driver problems; sorry. :/

Your .rpt files appear to be from 2.2_rc1. Upgrade to 2.2.1, please.

[cybersphinx]

Your .rpt files appear to be from 2.2_rc1. Upgrade to 2.2.1, please.

The later reports are from 2.2.1 (about 30% into the file), some of those are the map.h:225 thingy. Is that fixed in 2.2 yet?

[Zarel]

Yeah, the world_coord (map.h:225) one's been fixed. Some of the 2.2_rc1 ones have also been fixed, but most of them are pretty unfamiliar, and I can't figure out why they crashed. Ubuntu has so much better crash reports than Windows does... :/

I've fixed droid.c:4507 crash and the grp.c:221 crash. By my estimates, more than half of your crashes have been fixed for 2.2.2, nobby. Thanks for the useful crashdumps.

I don't understand the intSetStats crash (hci.c:5389), though. Thoughts?

[cybersphinx]

I don't understand the intSetStats crash (hci.c:5389), though. Thoughts?

Code: Select all

	psObj = intGetObject(id);
	if (psObj && psObj->selected)

The second line is 5389. Maybe the compiler inlined intGetObject here.

Code: Select all

static BASE_OBJECT *intGetObject(UDWORD id)
{
	BASE_OBJECT		*psObj;

	/* If this is a stats button, find the object button linked to it */
	if (id >= IDOBJ_STATSTART && id <= IDOBJ_STATEND)
	{
		id = IDOBJ_OBJSTART + id - IDOBJ_STATSTART;
	}

	/* Find the object that the ID refers to */
	ASSERT((SDWORD)id - IDOBJ_OBJSTART >= 0 && (SDWORD)id - IDOBJ_OBJSTART < numObjects, "Invalid button ID %u", id);
	psObj = apsObjectList[id - IDOBJ_OBJSTART];

	return psObj;
}

Then we skip the assert in release builds and access an invalid apsObjectList[id - IDOBJ_OBJSTART]. If that's correct, then changing the assert to ASSERT_OR_RETURN should fix it.

[Zarel]

Oh, you're right. I've committed that fix, too. So now we've fixed 90% of your crashes, I'd say?

[nobby]

Well done. If only the support on my commercial software was this responsive ;-}

Nigel.

[whippersnapper]

Well done. If only the support on my commercial software was this responsive ;-}

Nigel.

Ain't that the truth.

Regards, whip
.

[Zarel]

Well done. If only the support on my commercial software was this responsive ;-}

Tell me about it. Apple is completely silent about the OpenGL problems we have on their Nvidia Macs...

[nobby]

From your post above it sounds like most of the issues weren't driver related.

i. If I was to set up a dev environment to build and test the current trunk on Windows would you recommend MSVC (I have Visual Studio) or MinGW for the build. My C is a bit rusty but I can probably muddle through

ii. The offer of the video cards is still open if anyone thinks it's worth dogfooding on a Quadro.

Nigel.

[Zarel]

From your post above it sounds like most of the issues weren't driver related.

Most of the issues in the crash dump weren't driver related. Maybe we have good Quadro support, maybe driver crashes don't get dumped properly. It's hard to say.

In any case, we have known driver problems with Mac OS X and Nvidia, but they won't tell us if they're going to fix it, or how we can work around it. :/

i. If I was to set up a dev environment to build and test the current trunk on Windows would you recommend MSVC (I have Visual Studio) or MinGW for the build. My C is a bit rusty but I can probably muddle through

MinGW. Aside from Buginator, who isn't here right now and uses a ton of hacks, no one's managed to get MSVC working in ages.

[cybersphinx]

Aside from Buginator, who isn't here right now and uses a ton of hacks, no one's managed to get MSVC working in ages.

What hacks? I thought he said it was quite easy, as long as you don't want to do static builds. But since Mingw is similar to Linux, you'll probably find more people able to help you with problems.

[Zarel]

What hacks? I thought he said it was quite easy, as long as you don't want to do static builds. But since Mingw is similar to Linux, you'll probably find more people able to help you with problems.

I remember someone saying there were hacks invovled. Either it seemingly only works with Visual Studio 2003; I've been trying to get it working in VS 2008 with no dice.

I'm still bitter that most importantly, we can't compile an installer on either MSVC or MinGW in Windows.