On 11/10/2010 03:20 PM, Alexander Nozdrin wrote:
> Hi Mats,
> thanks for working on this.
> I think, the patch is Ok to push.
> I'd emphasize in a CS comment, that documentation should be updated:
> file specified in the '--defaults-file' option will not be loaded
> anymore if conf_file (or MYSQL::options::my_cnf_file for clients) is
> an absolute file path.
Well... it doesn't affect the external interface.
The difference is just that in subsequent calls, used internally when
loading the auto.cnf file and when loading a plugin, the default file
used at startup will be read unless you give an absolute file name to
the function (this is just done when the auto.cnf file is read
currently). Before this patch, the original default file was *always*
read, even ẃhen an absolute path name was given to the function,
contrary to the specification.
If an absolute file name is given to --defaults-file, it will be read
just fine, as it should be.
Just my few cents,
> On 10.11.2010 14:48, Mats Kindahl wrote:
>> #At file:///home/bzr/bugs/b57108-trunk-fixing/ based on
>> 3330 Mats Kindahl 2010-11-10
>> Bug #58109: Function my_load_defaults() does not read
>> correct file
>> When passed an absolute filename, my_load_defaults()
>> is supposed to read only that file, but instead it
>> reads values from the defaults file found in a previous
>> execution of the function (which is saved in
>> This patch fixes this by changing order of checks in
>> my_search_option_files() so that it first checks if an
>> absolute path was supplied, and in that case uses that
>> file only.