#1088 Debug cart returns incorrect values in console mode

[William McCabe]

Most SID tests fail with the testbench scripts in Windows 10. When the -console option is removed, the tests pass as expected. Below is a command line example of the error.

• The first command has the -console included which makes the test fail.
• The second command has the -console option removed and the test passes.
  I've tried this on 2 seperate computers with a Windows 10 environment.

Microsoft Windows [Version 10.0.14393]
(c) 2016 Microsoft Corporation. All rights reserved.

C:\Users\William>C:\Users\William\Documents\GTK3VICE-3.3-win32-r35872\x64sc -default -model c64c -VICIIfilter 0 -VICIIextpal -VICIIpalette pepto-pal.vpl -warp -debugcart -console -sidenginemodel 256 -limitcycles 53000000 G:\testprogs\SID\waveforms\waveforms-10-6581.prg
joy-win32.c:601: warning: function joy_arch_cmdline_options_init() is not implemented yet, continuing
DBGCART: exit(255) cycles elapsed: 49945501

C:\Users\William>C:\Users\William\Documents\GTK3VICE-3.3-win32-r35872\x64sc -default -model c64c -VICIIfilter 0 -VICIIextpal -VICIIpalette pepto-pal.vpl -warp -debugcart -sidenginemodel 256 -limitcycles 53000000 G:\testprogs\SID\waveforms\waveforms-10-6581.prg
joy-win32.c:601: warning: function joy_arch_cmdline_options_init() is not implemented yet, continuing
DBGCART: exit(0) cycles elapsed: 49924820

In addition to the above, screenshots are including the grey dot which causes some testbench screenshot tests to fail. A test where this is obvious is https://sourceforge.net/p/vice-emu/code/HEAD/tree/testprogs/VICII/rasterirq/rasterirq_hold.prg

[William McCabe]

The ticket was created as soon as I attached the screenshot which I wasn't ready to do. I wanted to change the title to be more generic because the rasterirq_hold test is not using console mode.

[gpz]

we did some tests... the wrong exitcode issue seems to exist only on windows, and only using the GTK version (it works with the SDL build...). which is weird. really :) Very likely its the test that is actually failing in this case. which makes no sense. shrug

as for the screenshots, there are different reference screenshots in the testbench for different VICs() - the testbench script picks the right one - so when you run
$ ./testbench.sh x64sc rasterirq
it just works :)
../VICII/rasterirq/ rasterirqhold.prg - ok

(i am guessing you are implementing support for VICE in your frontend? :))

[William McCabe]

Thanks, I've been testing with Ubuntu 18.10 and Windows 10 and only experience the issue with the Windows GTX version as you pointed out.

Yes, I'm currently working on a standalone version the frontend that I've succesfully added full support for VICE and HOXS64 in the unreleased version. I will be adding support for other emulators as well. It should be ready to release in the next few days. It will be a major update to the exisiting frontend and will include all functionality of your scripts including creating the html test results. In light of your response regarding the various screenshots I will make sure the frontend supports them as well. :)

[gpz]

should be fixed in r35897 :)

[William McCabe]

I've just confirmed r35897 fixes this bug. Ticket can be closed. Cheers. :)

[Strobe]

Confirmed fixed by willymanilly