[Crash-utility] resolve symbols: sial.so vs eppic.so

[Luc Chouinard]

Pulled trigger too quick.
Here's the variable you should set.

export EPPIC_LEGACY_MODE=1

If a more granular way of switching from old to new mode is needed, we
should discuss it.

   -Luc

> -----Original Message-----
> From: Luc Chouinard
> Sent: Thursday, September 06, 2012 9:01 AM
> To: 'Sebastian Ott'; Dave Anderson
> Cc: crash-utility at redhat.com; holzheu at linux.vnet.ibm.com
> Subject: RE: resolve symbols: sial.so vs eppic.so
> 
> Eppic now resolves symbols to their real typed instances directly.
> No need to know and cast from symbol address anymore.
> For backward compatibility, you can define environment variable
> 
> > -----Original Message-----
> > From: Sebastian Ott [mailto:sebott at linux.vnet.ibm.com]
> > Sent: Thursday, September 06, 2012 6:50 AM
> > To: Luc Chouinard; Dave Anderson
> > Cc: crash-utility at redhat.com; holzheu at linux.vnet.ibm.com
> > Subject: resolve symbols: sial.so vs eppic.so
> >
> > The way symbols are resolved in sial scripts is different between
> > sial.so and eppic.so. A symbol resolves to the address of the symbol
> > using sial.so and to the memory content using eppic.so.
> >
> > In crash 6.0.9 I got e.g.:
> >
> > crash> sym dev_kobj
> > 979138 (b) dev_kobj
> > crash> rd dev_kobj
> >           979138:  0000000174c9aac0                    ....t...
> > crash> dev_kobj
> > dev_kobj = $3 = (struct kobject *) 0x174c9aac0
> >
> > Using a sial script:
> >
> > crash> !cat foo.sial
> > string foo_help()  { return ""; }
> > string foo_opt()   { return ""; }
> > string foo_usage() { return ""; }
> >
> > int foo()
> > {
> >   printf("0x%x\n", dev_kobj);
> >
> >   return 1;
> > }
> >
> > ./eppic.so: shared object loaded
> > crash> load foo.sial
> > crash> foo
> > 0x74c9aac0
> >
> > ./sial.so: shared object loaded
> > crash> load foo.sial
> > crash> foo
> > 0x979138
> >
> > Is this an intentional change or a bug?
> >
> > Regards,
> > Sebastian