[Cocci] stripping renders spatch unusable [was: Starter issues with spatch]

[Julia Lawall]

On Thu, 25 Mar 2010, Jan Engelhardt wrote:

> Hi,
> 
> 
> On Thursday 2010-03-25 02:00, Jan Engelhardt wrote:
> >
> >I have a trivial issue trying to get started on spatch.
> >For a start, the manpage states:
> >
> >EXAMPLES
> >	[...]
> >
> >         ./spatch -sp_file foo.cocci foo.c -o /tmp/newfoo.c
> >
> >However, spatch (from cocc 0.2.2) does not seem to know any options. It 
> >does not support -h/--help either, so I can't tell if it's just the 
> >manpage being outdated.
> >
> >$ spatch -sp_file pfamily.smpl x.c -o new.c
> >Unknown option -sp_file.
> 
> 
> I guess I already figured it out:
> 
> 02:08 borg:../coccinelle-0.2.2/x > ./spatch -sp_file parfamily x.c
> init_defs_builtins: /usr/share/coccinelle/standard.h
> HANDLING: x.c
> No matches found for ->family .family
> Skipping:x.c
> 02:08 borg:../coccinelle-0.2.2/x > strip -s spatch
> 02:08 borg:../coccinelle-0.2.2/x > ./spatch -sp_file parfamily x.c
> Unknown option -sp_file.
> 
> 
> But why spatch relies on a symbol table is a bit beyond me.
> 
> Since many distros run strip just before packaging .rpm or similar, 
> there must be a way for spatch to work without the debugging symbol 
> table. (The ELF table is still available, readable with for example 
> `readelf -s spatch`.)

I found a few references saying that ocaml code compiled with -custom 
should not be stripped.  One place was in the Fedora packaging guidelines, 
so I guess the people who make distributions know what to do.  On the 
other hand, they said that stripping causes a no bytecode error, which is 
not what you are getting, so I'm not completely sure it is the same 
problem.

julia