Peter.Chubb@data61.csiro.au writes:
"Robert" == Robert P J Day rpjday@crashcourse.ca writes:
Robert> from lib/hashtable.c:
Robert> typedef struct _ENTRY { int used; ENTRY entry; } _ENTRY;
Robert> ok, that's just kind of creepy ... defining a typedef over top Robert> of a struct of the same name. does anyone else find that Robert> strange?
It's standard practice in some C styles. Personally I'd delete the struct tag, as the typedef should be used everywhere instead.
I'd rather drop the typedef. The struct/typedef is only referenced by name in two places.
First in include/search.h:
/* Opaque type for internal use. */ struct _ENTRY; [...] /* Data type for reentrant functions. */ struct hsearch_data { struct _ENTRY *table;
It is standard practice to use "struct foo" in such constructs.
The second use is further down in lib/hashtable.c:
htab->table = (_ENTRY *) calloc(htab->size + 1, sizeof(_ENTRY));
This line is terrible for two reasons: 1. Type-casting the return value of calloc is always wrong. 2. It is safer to use sizeof(*htab->table) instead of an explicit type.
If someone *really* wants to "fix" this, the proper thing is to drop the typedef and rewrite the calloc line to not explicitly mention the type.
Then again, this is code borrowed from uclibc/glibc, so perhaps it's best to simply leave it alone.