Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2009-02-19T01:32:44Z

That's a new one. I'd first see if the callback pointer is getting trashed by something. I'd suggest adding in SpTraceVcdC.cpp ::addCallback a printf("%p",fullcb) statement to then likewise in ::dumpFull on cip and cip->m_fullCb right before calling the callback. If the value changes something is trashing it, and you can try a GDB watchpoint on that address.

If there's multiple callbacks that get added, you can try commenting out the various calls into addCallback in the Verilated code to see if it's a specific one.

Original Redmine Comment
Author Name: Ralf Karge
Original Date: 2009-02-19T04:56:30Z

It seems that there is only one callback installed and that callback pointer appears correct.

SpTraceVcd::addCallback fullcb: 0x5f1212
SpTraceVcd::dump enter
SpTraceVcd::dumpFull enter
SpTraceVcd::dumpFull prep done. m_callbacks.size()=1
SpTraceVcd::dumpFull before callback - cip: 0x6fb20e0
(cip->m_fullcb) (this, cip->m_userthis, cip->m_code): (0x5f1212)(0x6fb4c80, 6fb1ed0, 1)
4 [main] Sim3285 3804 _cygtls::handle_exceptions: Error while dumping state (probably corrupted stack)
Segmentation fault (core dumped)

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2009-02-19T14:19:26Z

Hmm, are you close to out of memory?

I assume the callback is getting correctly invoked then, and you did build with '-O0 -ggdb'. Edit the traceFull callback to comment out a big block of code, and play the bisection game.

It might just be whatever first signal that calls vcdp->full..., so try commenting all but that first full call first.

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2009-05-08T18:50:21Z

I assume you got this figured out; if not please reopen.