[antlr-interest] SEGV when running AST builder in C target, antlr3.0

[Hardy, Stephen]

Jim,

I'm hacking the source to work around it.  Seems that testing for "-1"
as well as null is working, but there are several more places to get
through.

It's not really a complex problem; it happens right away.  

Anyway, don't worry about it for now - go have that beer...

Regards,
SJH


> -----Original Message-----
> From: Jim Idle [mailto:jimi at temporal-wave.com] 
> Sent: Tuesday, July 03, 2007 4:24 PM
> To: Hardy, Stephen; antlr-interest at antlr.org
> Subject: RE: [antlr-interest] SEGV when running AST builder 
> in C target, antlr3.0
> 
> Not seen that issue and I have some pretty complicated tree building
> parsers. However this may be related to a a fix made by a 
> recent change
> to ensure that pointers are initialized to NULL at rule 
> starts. Perhaps
> that also needs to be assured for the AST.stg too. The changes to ATS
> building were a bit close to the release day for comfort really ;-).
> 
> Jim
> 
> > -----Original Message-----
> > From: antlr-interest-bounces at antlr.org [mailto:antlr-interest-
> > bounces at antlr.org] On Behalf Of Hardy, Stephen
> > Sent: Tuesday, July 03, 2007 4:11 PM
> > To: antlr-interest at antlr.org
> > Subject: [antlr-interest] SEGV when running AST builder in C target,
> > antlr3.0
> > 
> > Jim,
> > 
> > when adding output=AST stuff to the C grammar, the runtime library
> gets
> > a seg fault in becomeRoot().  The main problem seems to be that the
> > code
> > tests pointers for NULL, whereas the pointers are actually getting
> > small
> > negative numbers because of various consitions in the token stream
> > (such
> > as empty).  Probably, there needs to be a macro like
> > IS_NOT_A_POINTER_OR_IS_NULL(newRootTree) etc.
> > 
> > Regards,
> > SJH
>