On 20 April 2014 22:38, Stefan Fuhrmann <stefan.fuhrm...@wandisco.com> wrote: > > On Sat, Apr 19, 2014 at 10:32 PM, Ivan Zhakov <i...@visualsvn.com> wrote: >> >> On 25 October 2013 15:12, <stef...@apache.org> wrote: >> > Author: stefan2 >> > Date: Fri Oct 25 11:12:27 2013 >> > New Revision: 1535686 >> > >> > URL: http://svn.apache.org/r1535686 >> > Log: >> > On the log-addressing branch: Automatically upgrade transactions from >> > physical to logical addressing mode for recently upgraded repositories. >> > Provide a test-case. >> > >> Hi Stefan, >> >> >> [...] >> >> > +/* Determine the checksum for the SIZE bytes of data starting at START >> > + * in FILE and return the result in *FNV1_CHECKSUM. >> > + * Use POOL for tempoary allocations. >> > + */ >> > +static svn_error_t * >> > +fnv1a_checksum_on_file_range(apr_uint32_t *fnv1_checksum, >> > + apr_file_t *file, >> > + apr_off_t start, >> > + apr_off_t size, >> > + apr_pool_t *pool) >> > +{ >> > + char buffer[4096]; >> > + >> Why you're using non-standard sized buffer for IO operations on stack? >> It should be apr_palloc(SVN__STREAM_CHUNK_SIZE). Is not it? > > > No need to use dynamically alloc the buffer > but not using SVN__STREAM_CHUNK_SIZE > as clearly an oversight. > Your reply doesn't have any arguments why no need to use buffer from pool like we do in all other places. Your fix looks wrong for me.
-- Ivan Zhakov CTO | VisualSVN | http://www.visualsvn.com
