> "Peter N. Lundblad" <peter@famlundblad.se> writes:
>>On Mon, 13 Jun 2005, Chris Foote wrote:
>>>Branko Čibej wrote:
>>
>>>>>+svn_info_t *
>>>>>+svn_info_dup (const svn_info_t *info, apr_pool_t *pool)
>>>>>+{
>>>>>+ svn_info_t *dupinfo = apr_pcalloc (pool, sizeof(*dupinfo));
>>>>>+
>>>>>+ /* Perform a trivial copy ... */
>>>>>+ *dupinfo = *info;
>>>>
>>>>Since you're copying the struct anyway, there's no sense in using
>>>>apr_pcalloc.
>>>
>>>True, although this is what svn_wc_entry_dup uses.
>>
>>I don't mind changing that too, even if this is really cycle-counting if
>>anything...
It's not cycle counting, because I don't care whether apr_palloc is faster than
apr_pcalloc. The reason to use apr_palloc is code clarity. It avoids the
reader wasting time checking: "This code seems to initialize the memory twice;
am I misunderstanding part of it? apr_pcalloc... checking... yes, that really
does initialise it. The copy "*dupinfo = *info"... checking... yes, that
really is copying the whole size of the structure, so that does too. Oh well."
svn_wc_entry_dup() fixed in r16647. Thanks for pointing it out.
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Oct 11 02:21:38 2005