Re: [PATCH v2 01/11] builtin/pack-objects.c: change check_pbase_path() to use ALLOC_GROW()

On Fri, Feb 28, 2014 at 9:20 PM, Michael Haggerty [off-list ref] wrote:

Duy,

The example in Documentation/technical/api-allocation-growing.txt does
it the same way as Dmitry:

    ALLOC_GROW(item, nr + 1, alloc);
    item[nr++] = value you like;

The alternative,

    nr++;
    ALLOC_GROW(item, nr, alloc);
    item[nr] = value you like;

is an extra line, which is at least a small argument for the variant
shown in the docs.  (Since ALLOC_GROW is a macro, it is not OK to use
"++nr" as its second argument.)  Personally, I also prefer the shorter
version.  The line

    item[nr++] = value

is an easy-to-recognize idiom, and

    ALLOC_GROW(item, nr + 1, alloc);

somehow makes it more transparent by how much more space will be needed.

So my vote is that the patches are OK the way Dmitry wrote them (mind, I
have only read through 05/11 so far).

I'm not saying all patches should do

nr++;
ALLOC_GROW(item, nr, alloc);

only those that do

if (..) realloc...;
nr++;
....

should be reordered. Those changes that do item[nr++] = yyy should be
kept. Anyway it's just an observation, not something that should block
these patches.
-- 
Duy