Re: More long-string woes

Tom Lane <tgl@sss.pgh.pa.us> writes:
> Joseph Barillari <jbarilla@princeton.edu> writes:
>> [ problems with long literals in plpgsql ]

> I'm amazed no one has noticed this before.  plpgsql_dstring_append
> is broken: it assumes it never needs to more than double the size
> of the string.

I've applied the attached patch to current CVS and the 7.2 branch
(it also works against 7.1 sources).  Seems to fix the cases you report.
        regards, tom lane

*** src/pl/plpgsql/src/pl_funcs.c.orig    Thu Nov 15 18:31:09 2001
--- src/pl/plpgsql/src/pl_funcs.c    Sun May  5 13:38:26 2002
***************
*** 64,69 ****
--- 64,70 ---- {     ds->value = palloc(ds->alloc = 512);     ds->used = 0;
+     ds->value[0] = '\0'; }  
***************
*** 86,95 **** plpgsql_dstring_append(PLpgSQL_dstring * ds, char *str) {     int            len = strlen(str); 
!     if (ds->used + len + 1 > ds->alloc)     {
!         ds->alloc *= 2;         ds->value = repalloc(ds->value, ds->alloc);     } 
--- 87,100 ---- plpgsql_dstring_append(PLpgSQL_dstring * ds, char *str) {     int            len = strlen(str);
+     int            needed = ds->used + len + 1; 
!     if (needed > ds->alloc)     {
!         /* might have to double more than once, if len is large */
!         do {
!             ds->alloc *= 2;
!         } while (needed > ds->alloc);         ds->value = repalloc(ds->value, ds->alloc);     }