"James Preston" <[EMAIL PROTECTED]> wrote in message
news:42442@palm-dev-forum...
> But the question remains, what is the "proper" way to address the
situation?
> Is the extraneous call to DmGetRecord the best/most efficient way?
In my opinion, yes. If DmResizeRecord didn't return a handle at all, this
would seem completely natural. There used to be multiple storage heaps, and
DmResizeRecord would move the record to a different heap if necessary.
That's why it returns a handle.
> Assuming
> that I know that I need to resize the record, and that I don't already
have
> a writable handle, is there a significant difference between:
>
> VoidHand nameRecHan = DmGetRecord(gNameDB, recIdx);
> nameRecHan = DmResizeRecord( gNameDB, recIdx, . . .
>
> and this:
>
> VoidHand nameRecHan = DmResizeRecord(gNameDB, recIdx, . . .
> nameRecHan = DmGetRecord(gNameDB, recIdx);
>
> If the resizing moves the record, the subsequent get should get the new
> handle, right? So the two cases should be equivalent. Is one to be
preferred
> for some reason, or is it just a wash?
They are, for all practical purposes, equivalent. However, the former is
preferred because resizing a record is considered an "edit", just like
DmWrite. One day, Palm OS may actually require the busy bit to be set before
a record is edited, either via DmWrite or via DmResizeRecord.
> In one of the previous threads on this topic (thanks, Danny, for the
> pointers to them), Jason Dawes included a code snippet in which he did a
> get, then a release before the resize. It seems to me that the intervening
> release is unnecessary (not to mention self-defeating for overcoming the
> assert in question). Doesn't DmReleaseRecord basically just unset the busy
> bit and optionally set the dirty bit?
You are correct. Here's Jason's code for reference:
> hRec = DmGetRecord(db, idx);
> if (MemHandleSize(hRec) != size_needed)
> {
> DmReleaseRecord(db, idx, false);
> hRec = DmResizeRecord(db, idx, size_needed);
> }
> ptr = MemHandleLock(hRec);
> // do stuff
> MemPtrUnlock(ptr);
> DmReleaseRecord(db, idx, modified);
On Palm OS 4.0 debug ROM's, the final call to DmReleaseRecord should
complain when the record was resized. I'd just get rid of the first call to
DmReleaseRecord.
--
Danny Epstein
OS Engineer, Palm Inc.
--
For information on using the Palm Developer Forums, or to unsubscribe, please see
http://www.palmos.com/dev/tech/support/forums/
