Re: [HACKERS] Questionable codes

Tatsuo Ishii <t-ishii(at)sra(dot)co(dot)jp> writes:
> I have found a few questionable codings. I'm not sure if it really
> hurts anything. Suggestions are welcome.

> 1) in storage/lmgr/lock.c: LockShmemSize()

> size += MAXALIGN(maxBackends * sizeof(PROC)); /* each MyProc */
> size += MAXALIGN(maxBackends * sizeof(LOCKMETHODCTL)); /* each

> shouldn't be:

> size += maxBackends * MAXALIGN(sizeof(PROC)); /* each MyProc */
> size += maxBackends * MAXALIGN(sizeof(LOCKMETHODCTL)); /* each

Probably, but I'm not sure it really makes any difference. We add on
10% or so slop after we've finished adding up all these numbers, anyway
;-)

> 2) in utils/hash/dynahash.c:hash_search():

> Assert(saveState.currElem && !(saveState.currElem = 0));

> Does anybody know what it is for?

That's part of that horribly ugly, non-reentrant HASH_REMOVE_SAVED
interface, isn't it? I have a to-do item to rip that code out and
replace it with a more reasonable design ... in the meantime, I don't
think it much matters whether the Assert could be tightened up ...

regards, tom lane