code for 1.9.19alphas

[Luke Kenneth Casson Leighton]

ok, further comments (this is posted to samba-technical, too).

instead of a separate free_xxx function, add an extra BOOL parameter to
make_xxxx.

if True, malloc.  if False, free.  call other make_xxxx sub-functions with
the extra BOOL parameter too.

reason: this will keep mallocs and frees in the same place.  this is
important so that future development doesn't get out of step (resulting in
loss of memory stuff).

there is a precedent for this already: the prs_xxx functions have a BOOL
io parameter.

if True, write into stream.  if False, read out of stream (possibly
vice-versa :-)

luke

> > Luke,
> >  
> > 	I intend to start hacking on your DCE/RPC code
> > in the main branch - just wanted to give you fair
> > warning :-).
> > 
> > Some of the things I intend to do :
> > 
> > 1). Tidy up the structures that say :
> > 
> > uint32  XXXX /* undocumented buffer pointer */
> > 
> > so that they get the address of the thing
> > they point to added into the rather than
> > the static '4' or '2' etc.
> 
> good idea.
>  
> > 2). Lots of the code looks like (an example
> > from svr_lsa.c):
> > 
> > --------------start fragment----------------------
> > /***************************************************************************
> > lsa_reply_lookup_sids
> >  ***************************************************************************/
> > static void lsa_reply_lookup_sids(prs_struct *rdata, 
> >                                 int num_entries, DOM_SID2 sid[MAX_LOOKUP_SIDS],
> >                                 char *dom_name, char *dom_sid,
> >                                 char *other_sid1, char *other_sid2, char *other_sid3)
> > {
> >         LSA_R_LOOKUP_SIDS r_l;
> >         DOM_R_REF ref;
> >         LSA_TRANS_NAME_ENUM names;
> >         uint32 mapped_count = 0;
> >  
> >         /* set up the LSA Lookup SIDs response */
> >         make_dom_ref(&ref, dom_name, dom_sid, other_sid1, other_sid2, other_sid3);
> >         make_lsa_trans_names(&names, num_entries, sid, &mapped_count);
> >         make_reply_lookup_sids(&r_l, &ref, &names, mapped_count, 0x0);
> > 
> >         /* store the response in the SMB stream */
> >         lsa_io_r_lookup_sids("", &r_l, rdata, 0);
> > }
> > 
> > ----------------end fragment---------------------------
> > Now most of the data is statically allocated (I'm thinking 
> > of the UNICODE strings stuff mainly.
> > 
> > What I indend to do is change it to something that looks 
> > like :
> > 
> > 
> > static void lsa_reply_lookup_sids(prs_struct *rdata, 
> >          int num_entries, DOM_SID2 sid[MAX_LOOKUP_SIDS],
> >          char *dom_name, char *dom_sid,
> >          char *other_sid1, char *other_sid2, char *other_sid3)
> > {
> >   LSA_R_LOOKUP_SIDS r_l;
> >   DOM_R_REF ref;
> >   LSA_TRANS_NAME_ENUM names;
> >   uint32 mapped_count = 0;
> >  
> >   /* set up the LSA Lookup SIDs response */
> >   make_dom_ref(&ref, dom_name, dom_sid, other_sid1, other_sid2, other_sid3);
> >   make_lsa_trans_names(&names, num_entries, sid, &mapped_count);
> >   make_reply_lookup_sids(&r_l, &ref, &names, mapped_count, 0x0);
> > 
> >   /* store the response in the SMB stream */
> >   lsa_io_r_lookup_sids("", &r_l, rdata, 0);
> > 
> > 
> >   /* Now free the stuff we allocated & linearized */
> >   free_dom_ref(....);
> >   free_lsa_trans_names(...);
> >   free_reply_lookup_sids(...);
> > }
> > 
> > Essentially - pair each 'make_xxxx' with a 'free_xxx'
> > and change the make_xxx() calls to do malloc's for
> > all the potentially large data.
> 
> the thought of doing this makes me go wobbly at the knees.
>  
> > This will get rid of the max unicode string stuff,
> > and make things dynamically allocated & freed
> > instead.
> 
> an alternative.  instead of a static structure, have a higher order
> function table.  in the example above, the functions would be
> 
> init_sid_table()
> set_sid_entry(); next_sid_entry();
> close_sid_table()
> 
> luke
>