Phucking hell! O_o;Urs wrote:CodeSonar finds bugs. Really. The pricing is awkward, from a phone call with their reseller I'd estimate a 5-digit sum in Euros for an unlimited number of seats. Depends on size of the project it seems.
Bug Friday - win & suffer with u-he!
- KVRAF
- 23102 posts since 7 Jan, 2009 from Croatia
-
- KVRAF
- 14658 posts since 19 Oct, 2003 from Berlin, Germany
So far, I didn't encounter any playback issues or random mutes with Cubase - at least nothing that triggers the "Bug Friday" tag.
I did have some findings however.
Tested on an i7 920, Win 7 Pro/Ultimate x64, RME HDSPe (currently on v3.38 with 256 samples as buffer), 6GB RAM, Cubase 6.5 x64, Cubase 7.06 (32bit), Cubase 7.06 x64. Test file was in 44/24, VST2 instances were used.
Creating a batch of 20 WAV channels with noise initially resulted in a "no processing" behavior until I saved and restarted.
I found out that creating a template in the 32bit environment doesn't load the 64x plugin automatically, neither is the plugin bridged.
Making a x64 version out of the 32bit version does work, but on saving and reloading, Cubase told me that it couldn't find Satin. Another change and save resulted in only the first two plugins (on the first two channels) missing. I reloaded and moved them to insert slot 2 each, rather than staying on insert slot 1 (rest of the project).
Randomly loading each host didn't result in muted instances. Just tried for the 25th time or so. No hickups or crackles either.
I did find out however, that the 32bit test project used up about 64% ASIO load in C7 32bit, and the x64 test project in both C6.5 and C7 x64 about 51%. That's over 10% less load for 20 instances.
Random mutes could still happen, but I really tried to push it with fast clicking around on the mouse and wild keyboard presses.
I did have some findings however.
Tested on an i7 920, Win 7 Pro/Ultimate x64, RME HDSPe (currently on v3.38 with 256 samples as buffer), 6GB RAM, Cubase 6.5 x64, Cubase 7.06 (32bit), Cubase 7.06 x64. Test file was in 44/24, VST2 instances were used.
Creating a batch of 20 WAV channels with noise initially resulted in a "no processing" behavior until I saved and restarted.
I found out that creating a template in the 32bit environment doesn't load the 64x plugin automatically, neither is the plugin bridged.
Making a x64 version out of the 32bit version does work, but on saving and reloading, Cubase told me that it couldn't find Satin. Another change and save resulted in only the first two plugins (on the first two channels) missing. I reloaded and moved them to insert slot 2 each, rather than staying on insert slot 1 (rest of the project).
Randomly loading each host didn't result in muted instances. Just tried for the 25th time or so. No hickups or crackles either.
I did find out however, that the 32bit test project used up about 64% ASIO load in C7 32bit, and the x64 test project in both C6.5 and C7 x64 about 51%. That's over 10% less load for 20 instances.
Random mutes could still happen, but I really tried to push it with fast clicking around on the mouse and wild keyboard presses.
Last edited by Compyfox on Wed Dec 04, 2013 4:50 pm, edited 2 times in total.
-
- KVRian
- 814 posts since 26 May, 2013 from France, Sisteron
What says the backtrace? Maybe you have a function where you forgot to return the value and with some optimisation, some value stays in eax, and without optimisation you have other value in eax. There is a gcc warning for that. Do you have -Wall -Wextra?Urs wrote:That's where it got weird: -O0 crashes AUValidation
I bet that you have initialisation or return value issue if the -Ox changes the behavior.
- u-he
- Topic Starter
- 28065 posts since 8 Aug, 2002 from Berlin
It doesn't really crash. It just issues a "ERROR: AudioUnitRender failed" with a cryptic error code...abique wrote:What says the backtrace? Maybe you have a function where you forgot to return the value and with some optimisation, some value stays in eax, and without optimisation you have other value in eax. There is a gcc warning for that. Do you have -Wall -Wextra?Urs wrote:That's where it got weird: -O0 crashes AUValidation
Good call about return value, but I haven't seen such warning lately. (Will look again)
- u-he
- Topic Starter
- 28065 posts since 8 Aug, 2002 from Berlin
Gotcha. That didn't trigger an error, but I eventually got a warning. *That* bug (AU only) is fixed!abique wrote:I think it is -Werror=return-type with gcc.
Thanks for the hint
#----
We're now waiting for our Windows build server to build through, then we'll have a new Satin version.
- u-he
- Topic Starter
- 28065 posts since 8 Aug, 2002 from Berlin
I'll sleep over it. Send me an emailabique wrote:Did I get a free plugin?
#---
Win builds are failing. Some server hiccup - I'm sure we'll have a new Satin by tomorrow!
-
- KVRer
- 13 posts since 23 Nov, 2005 from Moscow, Russia
clang static analyzer is for C++ too.
You need to ask clang to do static analysis, it's different than warnings — pass "--analyze" flag.
Enabling all warnings can be very useful too when hunting for bizzare bugs like this — add -Wall -Wextra in gcc. You can disable noisy warnings after that.
You need to ask clang to do static analysis, it's different than warnings — pass "--analyze" flag.
Enabling all warnings can be very useful too when hunting for bizzare bugs like this — add -Wall -Wextra in gcc. You can disable noisy warnings after that.
- Eugene 'HMage' Bujak.
-
- KVRist
- 172 posts since 20 Aug, 2010
Yes it would be but xcode or the compiler that comes with seems to ignore this option - can you confirm that? Because I tried setting that option before and it drove me crazy that it didn't work. The windows-compiler marks them as error most of the time but if it's mac-only code they are easy to miss...abique wrote:I think it is -Werror=return-type with gcc.
- u-he
- Topic Starter
- 28065 posts since 8 Aug, 2002 from Berlin
UPDATE: Final build, a very last chance! Less CPU hungry, focussed on the bug
Satin rev 2107 Win
Satin rev 2107 Mac
This one only has the Bug Friday log. If the debug code interfered with previous occurrances of The Bug, then we'll know. Otheriwse we will consider That bug fixed, and assume the existing appearances of muted Satins to be a different bug
Satin rev 2107 Win
Satin rev 2107 Mac
This one only has the Bug Friday log. If the debug code interfered with previous occurrances of The Bug, then we'll know. Otheriwse we will consider That bug fixed, and assume the existing appearances of muted Satins to be a different bug
- KVRAF
- 8620 posts since 2 Oct, 2006 from Leeds, UK
Urs do i overwrite the demo, i'm using it in a couple of projects so can i just replace it after testing?
Latest release and Socials: https://linktr.ee/ph.i.ltr3
-
- KVRAF
- 3817 posts since 8 Mar, 2006
- u-he
- Topic Starter
- 28065 posts since 8 Aug, 2002 from Berlin
You can, but please don't use it on existing projects. Due to some name changes within the plug-in you can't open settings saved with 1.1.1+ in earlier versions. It only works the other direction.musikmachine wrote:Urs do i overwrite the demo, i'm using it in a couple of projects so can i just replace it after testing?
- KVRAF
- 2138 posts since 8 Feb, 2007
I'm drooling over those numbers. my computer is not THAT inferior (9550) and according to DAWbench should be 25% slower which should - theoretically - give me at least 15 instances to play with.Compyfox wrote:I did have some findings however.
Tested on an i7 920, Win 7 Pro/Ultimate x64, RME HDSPe (currently on v3.38 with 256 samples as buffer), 6GB RAM, Cubase 6.5 x64, Cubase 7.06 (32bit), Cubase 7.06 x64. Test file was in 44/24, VST2 instances were used.
Creating a batch of 20 WAV channels with noise initially resulted in a "no processing" behavior until I saved and restarted.
I'd wish... with Rev 2107 I was able to reach 7 and the machine would freeze and halt (CPU consumption buildup
seem better. but overall its still not good).
Any new over that end ?