Horton

Thread

Commits

  1. Fix overridden callbacks in pg_rewind.

  1. [pg_rewind] use the passing callback instead of global function

    Junwang Zhao <zhjwpku@gmail.com> — 2023-04-26T01:51:13Z

    `local_traverse_files` and `libpq_traverse_files` both have a
callback parameter but instead use the global process_source_file
which is no good for function encapsulation.

-- 
Regards
Junwang Zhao

  2. Re: [pg_rewind] use the passing callback instead of global function

    Richard Guo <guofenglinux@gmail.com> — 2023-04-26T08:33:39Z

    On Wed, Apr 26, 2023 at 9:51 AM Junwang Zhao <zhjwpku@gmail.com> wrote:

> `local_traverse_files` and `libpq_traverse_files` both have a
> callback parameter but instead use the global process_source_file
> which is no good for function encapsulation.


Nice catch.  This should be a typo introduced by 37d2ff38.

While this patch is doing it correctly, I'm wondering that since both
kinds of source server (libpq and local) are using the same function
(i.e. process_source_file) to process source file list for
traverse_files operations, do we really need to provide a callback?  Or
will there be some kind of source server that may use different source
file processing function?

Thanks
Richard

  3. Re: [pg_rewind] use the passing callback instead of global function

    Daniel Gustafsson <daniel@yesql.se> — 2023-04-26T08:48:16Z

    > On 26 Apr 2023, at 10:33, Richard Guo <guofenglinux@gmail.com> wrote:
> 
> On Wed, Apr 26, 2023 at 9:51 AM Junwang Zhao <zhjwpku@gmail.com <mailto:zhjwpku@gmail.com>> wrote:
> `local_traverse_files` and `libpq_traverse_files` both have a
> callback parameter but instead use the global process_source_file
> which is no good for function encapsulation.
> 
> Nice catch.  This should be a typo introduced by 37d2ff38.

Agreed, I'll look at applying this after some testing.

> While this patch is doing it correctly, I'm wondering that since both
> kinds of source server (libpq and local) are using the same function
> (i.e. process_source_file) to process source file list for
> traverse_files operations, do we really need to provide a callback?  Or
> will there be some kind of source server that may use different source
> file processing function?


While there isn't one right now, removing the callback seems like imposing a
restriction that the refactoring in 37d2ff38 aimed to avoid.

--
Daniel Gustafsson