Looking through the postgresql source code, I notice that there are many places
were palloc is used but the return value is not checked to see if it is null.
There are a few places such as:
if (!PointerIsValid(result = palloc(CASH_BUFSZ + 2 - count + strlen(nsymbol)))) ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY), errmsg("out of memory")));
(taken from src/backend/utils/adt/cash.c), but at least within the that same
directory most occurrences of palloc are not checked.
Is this sloppy programming, or is there an automagical thing going on with
#defines that I'm just not seeing?
If it is just sloppy, perhaps we could use a new define in palloc.h, such as:
#define palloc_or_die(ptr,sz) \do { \ ptr = palloc(sz); \ if (!ptr) \ { \ ereport(ERROR, \
(errcode(ERRCODE_OUT_OF_MEMORY),\ errmsg("Out of memory"))); \ } \} while(0);
And then, in all places where the code does not currently check the return value
of palloc, the code could be changed to use palloc_or_die instead. Of course,
I'd be happy if someone has a better name for the macro, perhaps something more
brief?
I can go over the code "with a fine tooth comb" and replace the offending
occurrences. Does the community think this is a good idea?
mark