Alfranio Correia wrote:
> Hi,
>
> Can you take a look at this bug?
>
Sure.
> Besides, I would like to have your feedback on five topics:
>
> 1 - The documentation is not clear on what should happen when both
> options --slave-load-tmpdir
> and --secure-file-priv are defined.
> 1.1 In the patch, --secure-file-priv takes precedence over
> --slave-load-tmpdir whenever the LOAD_FILE and LOAD DATA INFILE are
> used. Is that ok?
>
I think this would confusing the user if he had both options defined,
and found out that --slave-load-tmpdir did not work as expected.
I's suggest to allow slave SQL thread to load files from the directory
specified by --slave-load-tmpdir too besides --secure-file-priv.
> 2 If one assigns invalid directories to any of these options, the slave
> crashes. This should
> be fixed before pushing the patch.
> 2.1 Should I do this when the options are initialized in the mysqld.cc?
> 2.2 Should I do this in the log_event.cc?
> 2.3 What should I do if the options are invalid?
I think 2.1 is a better choice, and I think if any of these options are
invalid, mysqld should fail with an error message to state that.
>
> 3. Sketch of the test:
> 3.1 Verify if the option is invalid:
> Transform the a relative path to an absolute path
> Check if the directory exists.
> Check if we have permissions to create and delete files.
>
> Cheers.
