Win10 upgrade breaking demos, yes or no?

category: general [glöplog]
Hey everybody,

I just made some hardware upgrades to the @party compo machine but it's still running Win7 32bit. I was about to upgrade it to Win7 64bit, but was wondering if anybody had any experience running demos on Win10 and what percentage (if any) of your favorite demos from the last decade broke in the Win7-Win10 upgrade.


- Dr.Claw
added on the 2016-05-02 21:53:57 by DrClaw DrClaw
hmm, didn't make any exhaustive tests but all windows demos seem to run fine on my windows 10 machine. some 4ks missing some runtime .dlls as usual, but still works, didn't notice much difference from 7 or 8 to 10.

there were some issues with textmode demos i believe, but those were already happening on windows 7.
added on the 2016-05-03 00:00:25 by psenough psenough
Yup, textmode has no native fullscreen anymore since Vista.
But even that got a solution floating around somewhere...someone coded a fullscreen-viewer for textmos.
I never had problems.
added on the 2016-05-03 11:25:13 by xernobyl xernobyl
The only problem I've had was that the functionality for blocking "unknown" executables seem to be even more eager than it was on Win 8.1.
Apart from that I haven't had any issues yet that I didn't already have with 8.1.

Win 10 seems to be quite good from my perspective.
added on the 2016-05-03 11:36:49 by lug00ber lug00ber
Yup, textmode has no native fullscreen anymore since Vista.

That being said in Win10 you can fullscreen a console again (Alt-Enter)
added on the 2016-05-03 12:52:59 by Gargaj Gargaj
no win10 issues so far either, at least not with prods that worked under win7/x64.
although im not exactly sure how it looks performance wise, sometimes I do have a feeling theres more background stuff going on in win10.
that or my good old 660 ti just doesnt cut it anymore :p
added on the 2016-05-03 13:03:18 by wysiwtf wysiwtf
How about music in 1k intros? https://www.pouet.net/prod.php?which=66038
added on the 2016-05-03 19:31:12 by yzi yzi
Once you have a solid Win10 install, it shouldn't be too different from Win7, in fact slightly better I think.

I like and use Win 7. Some (Windows) demos require a few dlls, which are in Win8+ (or you can keep Win 7 and just downloads those few dlls).

If you like the Win10 UX, for just running demo exes do update. For general use, note that there are quite a few threads about the Win7-Win10 nag non-clean install. (Suboptimal result, similar to Yosemite threads for Mac). The general advice to fix issues is a clean install.

My impression is that Win8 and 10 is Win7 not upgraded but looking at what they have and trimming it rather than adding features. It's Win7 with touch.
added on the 2016-05-04 00:03:25 by Photon Photon
I just did a fresh install of Win 10 yesterday and it took much less work to get things running smoothly than when I did a fresh Win 7 install last year. Plus I feel like with the right settings 10's UI is even better at getting out of your face, which I greatly prefer on a compo machine or presentation system.

Thanks for all the input!
added on the 2016-05-04 08:14:20 by DrClaw DrClaw
Photon: you can also just "Reset" the installation from win10 itself after you have updated from previous version AND have some quirky things, its a bit less of a hassle than installing from beginning again: http://windows.microsoft.com/en-us/windows-10/windows-10-recovery-options

I've been running it since first Insider release.
added on the 2016-05-04 08:31:08 by leGend leGend
Windows 8 causes crashes on Crinkler-compressed intros that use the Speech API. As Blueberry wrote:
Some essential functions in ole32.dll have changed to Forwarded RVA in Windows 8, so the Crinkler import code crashes.

It forwards to a DLL with a very long and weird name. I don't see much hope for SAPI in Crinkler-compressed intros on Windows 8+ (unless you do manual import by name), but we will look into it.
added on the 2016-05-04 09:52:12 by Seven Seven
So, Win10 ran fine on the @party compo machine. Though today I went to run "2nd Stage Boss" and it fails silently. It ran fine back in June on the same machine. Win10 64bit, i7, R9 Fury. I have noticed that "Escape Through Subspace 1k" also fails similarly if I run the 1k version, though running the "safe" 2k version works... hrm...
added on the 2016-10-22 07:30:59 by DrClaw DrClaw
just disable Windows Update for 7 =)
added on the 2016-10-22 08:30:07 by g0blinish g0blinish
Hi DrClaw,

Another possibility could be videocard driver updates. I couldn't find the compo machine spec at the Demospash site, are you using an Nvidia or AMD card?

There are several things changed between the safe and nonsafe versions of Escape Through Subspace 1K, and I'm suspecting one particular hack that caused me a day of pain at Assembly (a videocard driver change that ONLY happened for the 1080 GTX card). The fixed version worked on all cards, but maybe Nvidia changes something back? Tomorrow I can make some custom versions with different hacks removed, if you want to help finding out what causes this.

Just curious, does this intro works: http://www.pouet.net/prod.php?which=60880 ?
added on the 2016-10-22 08:46:12 by Seven Seven
And I just noticed you mentioned the specs... Although the R9 Fury lowers the probability that it is that particular hack that's problematic, the offer for debug versions still stands.
added on the 2016-10-22 08:48:43 by Seven Seven
In general I haven't noticed too many demos not working since I switched to win10 from win7. At least no more than usual incompatibility you sometimes get, for example a few demos didn't work on my win7 setup either. However haven't exactly gone back over every release to conduct an exhaustive test.

Out of curiosity and since I figure size restricted prods will be the most likely to have some issues I checked 2nd Stage Boss and it fails silently for me also on win10. I know it worked fine for me under win7.

I checked my old 1k releases and the first one I made http://www.pouet.net/prod.php?which=64235 runs just fine for me on win10. It was made on win7 and used the previous version of crinkler, 1.4 I think it was? I then tested this one http://www.pouet.net/prod.php?which=66496 which was made on the same win7 setup and is almost identical in source code besides the shader and music. It was packed using the new 2.0 crinkler using the tinyheader option. This fails silently on my win10 setup. Now the interesting part is I just made another 1k for TRSAC and again it's basically the same source code with new shader and music and packed with 2.0 crinkler/tinyheader option. It works on win10 and I received feedback that it also works on the compo machine which I understand is running win10.

To see if I could figure out the difference I tried doing crinkler /RECOMPRESS on the non working 1k, it still didn't work. I recompiled the original source without changing anything and it works on win10. Didn't seem to make any difference if I changed the platform toolset in VS between 120 and 140 either. Conclusion is I am not exactly sure what is changed by simply compiling it on win10 that makes it work as there seems to be nothing changed but I also am not an expert on low level stuff and dll loading and such which may be the issue.
added on the 2016-10-22 09:47:37 by drift drift
some directx demos seem to have issues after a recent nvidia drivers update though. e.g. last few excess demos have missing geometry and last few fairlight demos have weird aliasing artifacts or just buggy geometry (e.g. the 'numb res mask' in ziphead). even the more reason to stick to youtube! :D
added on the 2016-10-22 15:37:27 by Maali Maali
yea the nvidia drivers around the 1080 release introduced some bugs related to conditional loops and resource writing in compute shaders and they never fixed them, breaking our deferred lighting. (dear nvidia: contact me for repro)
added on the 2016-10-22 16:14:46 by smash smash
In my expericence older windows demos run generally better on windows 10 then they did on windows 7.
kkowboy by blasphemy&purple for example didn't run on my win7, but does again on win10
added on the 2016-10-22 16:18:35 by v3nom v3nom
@smash: interesting that you also traced this down to somehow relate to variable loops. That was my conclusion also, with the added bit that they seem to confuse the limitations of tex2D() and tex2Dlod() inside such loops.
added on the 2016-10-24 00:19:37 by kusma kusma
kusma: interesting. the breaking code in my case was a variable loop that distributed work across cs threads - that was based on a value read from a texture. i had to pad the loop to match thread count, like it was not properly obeying/setting the execution mask.
added on the 2016-10-24 00:43:37 by smash smash
If there's a question of getting a reproduction case to NVIDIA, shockingly I'm okay connected on that front, so just get in touch :)
added on the 2016-10-24 11:27:52 by gloom gloom
gloom: why on earth would we want to do something practical about this when we can moan about it on pouet?
added on the 2016-10-25 00:39:02 by smash smash
Hey Seven, if you want to send me some debug versions go right ahead. I'd like to get to the bottom of this as well so I know what to tell somebody when their compressed entry fails to run on the compo machine. drclaw AT dogsolitude DOT org. Thanks!
added on the 2016-10-25 18:32:13 by DrClaw DrClaw