Re: GST segfaults when running under GDB

On Sat, 2015-02-28 at 18:35 +0100, Holger Hans Peter Freyther wrote:
> On Sat, Feb 28, 2015 at 02:42:27PM +0000, Jan Vrany wrote:
> > Hi, 
> 
> Hi!
> 
> > Is it normal? If so, do I have to setup GDB specially or
> > compile GST specially to be able to debug it?
> 
> Yes that is normal. For the incremental/generational garbage
> collection one needs to know which objects changed since the
> last operation. The classic way is to have a read/write barrier
> (e.g. keep track everytime a pointer is followed to another
> object). The approach Paolo picked was to use mprotect on the
> pages the object is allocated.
> 
> This means that a SIGSEGV will be generated on these objects
> and the garbage collector remembers it.

Ah, yes, I know that. Stupid me, just did not think of it in
this case. Sorry! 

> 
> a.) In gdb you can do:
> 
> handle SIGSEGV nostop noprint
> b abort
> b exit
> 
> b.) You can compile GST with another GC and/or in libgst/oop.h
> force "NO_SIGSEGV_HANDLING"

I'd prefer the former than differently compiled VM.
Thanks a lot! 

Best, Jan