<takes deep breath>
I'm trying to be calm. Really. It's just hard- P.U.R.E.'s about ready to beta-test, I was so happy and ready to show it off, get new models done, etc. I don't want to be a dick about this- I don't understand 99% of what went into that patch, which is frustrating- I can't even begin to help figure out what's wrong.
Lemme check my archive of builds, and look at the last one I had before this patch was introduced...
Ok... the last build I installed before this patch was 5875. I'm going to test it now, to be absolutely sure that everything was working at that point.
<tests>
New Bugs, Rev 5879+
Moderator: Moderators
Re: New Bugs, Rev 5879+
If need be, I can pretty quickly make you a series of builds so you can find the exact rev, and then try to split that rev up. I don't think you have a compiler configured?
Re: New Bugs, Rev 5879+
I haven't compiled Spring in ages, no. I got tired of maintaining SVN, when I wasn't really doing much in the way of engine coding, and got rid of everything.
I've just tested 5875... everything is working at that point. CEGs draw normally, no serious rendering errors, no graphical corruption after /give all, etc. Framerate is down a little bit, but I dunno whether that is Tobi's optimization or not- that helped terrifically with late-game performance, and on the World Builder demo, where so much crap is happening.
I've just tested 5875... everything is working at that point. CEGs draw normally, no serious rendering errors, no graphical corruption after /give all, etc. Framerate is down a little bit, but I dunno whether that is Tobi's optimization or not- that helped terrifically with late-game performance, and on the World Builder demo, where so much crap is happening.
Last edited by Argh on 21 May 2008, 00:36, edited 1 time in total.
Re: New Bugs, Rev 5879+
Yay! A heated dispute over an important issue is being resolved without moderation-mayhem and BAAAAW implosion, but instead with taking a step back and re-examining the situation! Let's have a group hug!
Re: New Bugs, Rev 5879+
numbers say more than thousand words 
:
r5870: 95fps
r5875: 57fps (!!!)
r5879: 75fps (that's the commit with my rendering.patch!)
r5879(CompressTextures=1): 30fps
r5850+mypatch: ~100-110fps (CompressTextures=1 has no influence!)
:r5870: 95fps
r5875: 57fps (!!!)
r5879: 75fps (that's the commit with my rendering.patch!)
r5879(CompressTextures=1): 30fps
r5850+mypatch: ~100-110fps (CompressTextures=1 has no influence!)
Re: New Bugs, Rev 5879+
This isn't about speed, on your chipset. This is about things not working correctly at all. If it ran half as fast, but everything rendered correctly, it'd still be better than something that only works on 10% of the chipsets players use.
If you want to send me a Windows compile of the EXE, of 5850 + your patch, I can test it. If it's still broken, then I guess that tells us what we want to know.
If you want to send me a Windows compile of the EXE, of 5850 + your patch, I can test it. If it's still broken, then I guess that tells us what we want to know.
Re: New Bugs, Rev 5879+
All right, lemme test it, see what happens...
Re: New Bugs, Rev 5879+
Fight my war.
Re: New Bugs, Rev 5879+
Test results:
5850 + patch:
1. UsePBO = 1, CompressTextures = 0:
All CEGs draw normally. No corruption seen in graphical elements. UI draws reliably. My music Gadget had to be disabled first, before it would run the UI, because it's using a feature that was fixed and working after 5850, but caused problems with the VFS prior to that fix.
Framerate is as good as I've ever seen in Spring.
IOW, everything works perfectly. That's really, really interesting.
I'm going to test the rest of the UsePBO / CompressTextures combos now, but that pretty much settles that- between 5850 and 5879, something changed.
Either that, or it's a compiler error that's causing this, in the Buildbot builds. You sent a 10MB EXE- I assume with debugging symbols?
5850 + patch:
1. UsePBO = 1, CompressTextures = 0:
All CEGs draw normally. No corruption seen in graphical elements. UI draws reliably. My music Gadget had to be disabled first, before it would run the UI, because it's using a feature that was fixed and working after 5850, but caused problems with the VFS prior to that fix.
Framerate is as good as I've ever seen in Spring.
IOW, everything works perfectly. That's really, really interesting.
I'm going to test the rest of the UsePBO / CompressTextures combos now, but that pretty much settles that- between 5850 and 5879, something changed.
Either that, or it's a compiler error that's causing this, in the Buildbot builds. You sent a 10MB EXE- I assume with debugging symbols?
Re: New Bugs, Rev 5879+
2. UsePBO = 0, CompressTextures = 0:
All elements draw correctly. No corruption seen. UI draws correctly.
Performance is slower than with UsePBO.
Hmm. Time to look at the 29 patches between the two. I don't even see a point in testing further- it works as advertised, the root problem is obviously not in jK's patch.
All elements draw correctly. No corruption seen. UI draws correctly.
Performance is slower than with UsePBO.
Hmm. Time to look at the 29 patches between the two. I don't even see a point in testing further- it works as advertised, the root problem is obviously not in jK's patch.
Re: New Bugs, Rev 5879+
I'll look around, and if nothing obvious shows up on my computer, I'll upload a solid archive with 30 spring exes.
Re: New Bugs, Rev 5879+
After looking at the list, it appears that Tobi applied the patch in 3 parts, and I admit, I didn't test them all- I was out of town when they showed up.
I'm betting 5877 is the culprit, given the symptoms.
I'm betting 5877 is the culprit, given the symptoms.
-
[Krogoth86]
- Posts: 1176
- Joined: 23 Aug 2007, 19:46
Re: New Bugs, Rev 5879+
I don't know if you already know but I recently tried the SVN5917 and got very heavy performance problems...
I.e. this is only about the standard non-Ice-UI buildmenu (tried it with MA) because when it shows up the framrate goes down to like 2 and even when I host I start to get high pings. The buildpics also have some visual errors and the highlighting doesn't work correctly i.e. when your cursor over a very right buildpic the "edge highlight" will appear on the very left button on row below. When you deselect the unit everything turns back to normal...
Didn't have any problems like this in some previous SVNs although I can't recall the exact versions I tried out...
I.e. this is only about the standard non-Ice-UI buildmenu (tried it with MA) because when it shows up the framrate goes down to like 2 and even when I host I start to get high pings. The buildpics also have some visual errors and the highlighting doesn't work correctly i.e. when your cursor over a very right buildpic the "edge highlight" will appear on the very left button on row below. When you deselect the unit everything turns back to normal...
Didn't have any problems like this in some previous SVNs although I can't recall the exact versions I tried out...
Re: New Bugs, Rev 5879+
Terrain is toast here too BTW, it used to work fine before the big patch.
Re: New Bugs, Rev 5879+
I read some posts and get the feeling you're lacking for data, so I'm going to provide some - hope it comes in handy.
Testing the latest version from the link Argh posted, specifically spring_0.76b1+svn5917.exe , aside from the fullscreen only alt-tab, I havn't found any issues. Indeed, it was easier to start than the normal spring, I didn't have to set this new exe to single-core whatever the heck to get it to run above 5fps - not sure whether that's a bugfix or just it recognizing the exe is called the same thing.
Anyway - I'm assuming sands of war is meant to be that horrible, everything is washed out mucky yellow colour? All of the other maps look fine. I can't see anything different about the water, does it need a specific map?
I tested it with no AI scripts or anything, just a commander with which I built some stuff, with both BA and XTA as provided in the package.
Tree shadows rapidly turn into big squares as you zoom out, as do smoke/particle shadows, but I'm guessing that's intended behavior as well. I have my normal kinda-cruddy fps, usually hovering around 20-30 once there's a 100+ units, same as normal spring.
My system stats, as reported by DXDiag:
Laptop
Intel core2 Duo T7500 @2.20GHz (2 cpus) ~2.2GHz
Memory: 2046MB RAM
Graphic chip:
NVIDEA GeForce 8600M GS
Approx total memory: 1008MB (It's supposed to be a gig. I dunno where the rest is ^_^)
Driver: nvd3dum.dll
Version: 7.15.0011.0108 - the date below tells me my graphics drivers are almost a year out of date now, I should indeed fix that ( 05/04/2007 )
Sound:
Realtek High Definition Audio
I was running it on windows vista home basic, in windowed mode, in 900x675 resolution. It ran about as fast as spring usually does with graphics settings to prettyish, and I had water on Reflective + Refractive. It looked identical on delta siege.
In fullscreen, it worked fine (same speed) until alt-tabbing, at which point I got horrible graphical corruption everywhere - observe.
I didn't install over a spring install, but I did install side by side (different directory same user) so I expect it took several settings from that.
I hope this is helpful to someone, as there don't seem to be many other tests - I was trying to look at the water too but I'm guessing it's either map or a different build specific. Either way, more data is presumably helpful.
One final thing - there was odd grey shadowed text which seemed to be displaying corrupt incorrect strings behind the annoying press F11 to go away eyes script. Everything else, including all of the lua shinies that I tried (not all of them, but a random selection), worked perfectly. It wasn't a major bug, but there were similar effects behind a couple of numbers in XTA's custom costs popup - they didn't get in the way or seem to be spreading or anything.
I think it's worth noting that although the graphical corruption bug is not present in normal spring, if you alt-tab in fullscreen often enough it will eventually just crash on this system, around 3-4 times usually does it, but I think it's random and not something that accumulates. This is most of the reason why I play in windowed mode. Finally, sorry for the huge post.
Testing the latest version from the link Argh posted, specifically spring_0.76b1+svn5917.exe , aside from the fullscreen only alt-tab, I havn't found any issues. Indeed, it was easier to start than the normal spring, I didn't have to set this new exe to single-core whatever the heck to get it to run above 5fps - not sure whether that's a bugfix or just it recognizing the exe is called the same thing.
Anyway - I'm assuming sands of war is meant to be that horrible, everything is washed out mucky yellow colour? All of the other maps look fine. I can't see anything different about the water, does it need a specific map?
I tested it with no AI scripts or anything, just a commander with which I built some stuff, with both BA and XTA as provided in the package.
Tree shadows rapidly turn into big squares as you zoom out, as do smoke/particle shadows, but I'm guessing that's intended behavior as well. I have my normal kinda-cruddy fps, usually hovering around 20-30 once there's a 100+ units, same as normal spring.
My system stats, as reported by DXDiag:
Laptop
Intel core2 Duo T7500 @2.20GHz (2 cpus) ~2.2GHz
Memory: 2046MB RAM
Graphic chip:
NVIDEA GeForce 8600M GS
Approx total memory: 1008MB (It's supposed to be a gig. I dunno where the rest is ^_^)
Driver: nvd3dum.dll
Version: 7.15.0011.0108 - the date below tells me my graphics drivers are almost a year out of date now, I should indeed fix that ( 05/04/2007 )
Sound:
Realtek High Definition Audio
I was running it on windows vista home basic, in windowed mode, in 900x675 resolution. It ran about as fast as spring usually does with graphics settings to prettyish, and I had water on Reflective + Refractive. It looked identical on delta siege.
In fullscreen, it worked fine (same speed) until alt-tabbing, at which point I got horrible graphical corruption everywhere - observe.
I didn't install over a spring install, but I did install side by side (different directory same user) so I expect it took several settings from that.
I hope this is helpful to someone, as there don't seem to be many other tests - I was trying to look at the water too but I'm guessing it's either map or a different build specific. Either way, more data is presumably helpful.
One final thing - there was odd grey shadowed text which seemed to be displaying corrupt incorrect strings behind the annoying press F11 to go away eyes script. Everything else, including all of the lua shinies that I tried (not all of them, but a random selection), worked perfectly. It wasn't a major bug, but there were similar effects behind a couple of numbers in XTA's custom costs popup - they didn't get in the way or seem to be spreading or anything.
I think it's worth noting that although the graphical corruption bug is not present in normal spring, if you alt-tab in fullscreen often enough it will eventually just crash on this system, around 3-4 times usually does it, but I think it's random and not something that accumulates. This is most of the reason why I play in windowed mode. Finally, sorry for the huge post.
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: New Bugs, Rev 5879+
Is buildbot working again? Cause if So point me in a direction and I'll do some testing here as well.
Re: New Bugs, Rev 5879+
no need to, i already know which commit is broken, i only don't know why, so i need to compile it myself ...
Re: New Bugs, Rev 5879+
@jK: Excellent! We're finally getting somewhere with this 
Which commit btw?
@ForbodingAngel:
Buildbot's working. Go here, to get the latest binaries:
http://planetspring.free.fr/spring/installer/
If you want to play around with Buildbot, and test various revisions, we can do so now, thanks to Bibim. Just join the #buildserv channel online, then you can use !rebuild, or !rebuild "number-of-revision-here" to build specific versions of Spring.
Which commit btw?@ForbodingAngel:
Buildbot's working. Go here, to get the latest binaries:
http://planetspring.free.fr/spring/installer/
If you want to play around with Buildbot, and test various revisions, we can do so now, thanks to Bibim. Just join the #buildserv channel online, then you can use !rebuild, or !rebuild "number-of-revision-here" to build specific versions of Spring.