crash when opening a ngspice wbk from v6

Description

Steps to reproduce

1. open a simulation schematic comming from V6.0
2. Click [Inspect][Simulator...]
3. In the simulator window, click [File][Open Workbook...]
4. crash

KiCad Version

Application: KiCad x64 on x64

Version: (7.0.0-rc1-253-g7844602210), release build

Libraries:
	wxWidgets 3.2.1
	FreeType 2.12.1
	HarfBuzz 5.0.1
	FontConfig 2.14.1

Platform: Windows 11 (build 22621), 64-bit edition, 64 bit, Little endian, wxMSW

	wxWidgets: 3.2.1 (wchar_t,wx containers)
	Boost: 1.80.0
	OCC: 7.6.2
	Curl: 7.86.0-DEV
	ngspice: 38
	Compiler: Visual C++ 1934 without C++ ABI

Build settings:
	KICAD_SPICE=ON

added 7.0 priorityundecided statusnew labels

added prioritycritical simulation labels and removed priorityundecided label

assigned to @jeffyoung

changed milestone to %7.0

@KICAD_user can you upload a project which shows this? I tried "Sallen_Key" and "QEI", but they both worked fine (and most of my other simulation tests don't include workbooks).

Oh wait, you're not even getting to the Open File dialog? That sounds more like a menu issue. Can you get a stack-trace?

<Can you get a stack-trace?>

Sure, when you tell me how to do this?

Hmm... not on MSW.

@kicad/code/lead_devs can someone give Gerhard some MSW instructions for getting a back-trace?

(Alternatively, can anyone on MSW reproduce this?)

Here the project....

simul_V6.zip

The model parameter files are missing.

If sending a project, best is to add the model parameter libs to the project folder and attach them to the symbols accordingly.

Yeah, when opening that workbook I get two file-not-found errors, but still no crash.

Same on Linux. Two file not found messages. No crash.

Here the project with embedded models. I can reproduce the crash to 100%. Anon. reporting is activated.

simul_V6b.zip

I'm guessing this is KICAD-7P

I can reproduce it, yesterday's nightly, on Windows 10.

Still cannot find one of the simulation models on Linux. This doesn't look like a valid path to me.

models\BC546.lib is Windows speech. models/BC546.lib would be valid universially.

Even without being able to reproduce it, this code is fishy, we are indexing into C-style arrays unchecked in SIM_PLOT_PANEL::addTrace which is not great, so at minimum any caller of addTrace needs to bounds-check the aX and aY arrays.

Even fixing the file path issue, I still don't get a crash on Linux. The path separator is another bug. Is this possibly yet another Windows built with MSVC only bug? @charrasjp are you seeing this on your msys2 builds?

This crash is likely dependent on system-specific memory allocation / protection; the Sentry reports show an out-of-bounds access.

Note that our different platforms have different behaviors when accessing out-of-bounds; a crash is not guaranteed especially on Linux.

I agree that would should not blindly be attempting to access an array without bound checking just like we shouldn't blindly assume that pointers are valid.

Using simul_V6b.zip I am seeing the crash

can you break in SIM_PLOT_PANEL::addTrace and inspect the length of the arrays compared to the number of points specified?

I'm getting a simulation failure so I'm guessing there are no trace points. The question then would be why is this code being called when the simulation fails?

It seems to me that the wxASSERT_MSG( false, wxT( "Plot type missing AC_PHASE or AC_MAG bit" ) ); in SIM_PLOT_FRAME::updatePlot() needs to return, not break. Otherwise we get to addPlot with an empty data_y.

Is this assert firing for those who see it crashing?

ping

unassigned @jeffyoung

assigned to @jeffyoung

closed with commit 4e50c531

added statusfix-committed label and removed statusnew label

added statusfix-released label and removed statusfix-committed label