Обсуждение: pg_xlogdump follow into the future

On 2016-07-14 13:46:23 +0200, Magnus Hagander wrote:
> Currently, if you run pg_xlogdump with -f, you have to specify an end
> position in an existing file, or if you don't it will only follow until the
> end of the current file.

That's because specifying a file explicitly says that you only want to
look at that file, specifying two files that you want the range
inclusively between the two files.  -f works if you just use -s.


> I'd appreciate a review of that by someone who's done more work on the xlog
> stuff, but it seems trivial to me. Not sure I can argue it's a bugfix
> though, since the usecase simply did not work...

I'd say it's working as intended, and you want to change that
intent. That's fair, but I'd not call it a bug, and I'd say it's not
really 9.6 material.

> diff --git a/src/bin/pg_xlogdump/pg_xlogdump.c b/src/bin/pg_xlogdump/pg_xlogdump.c
> index c0a6816..2f1018b 100644
> --- a/src/bin/pg_xlogdump/pg_xlogdump.c
> +++ b/src/bin/pg_xlogdump/pg_xlogdump.c
> @@ -901,8 +901,14 @@ main(int argc, char **argv)
>              goto bad_argument;
>          }
>  
> -        /* no second file specified, set end position */
> -        if (!(optind + 1 < argc) && XLogRecPtrIsInvalid(private.endptr))
> +        /*
> +         * No second file specified, so unless we are in follow mode,
> +         * set the end position to the end of the same segment as
> +         * the start position.
> +         */
> +        if (!(optind + 1 < argc) &&
> +            XLogRecPtrIsInvalid(private.endptr) &&
> +            !config.follow)
>              XLogSegNoOffsetToRecPtr(segno + 1, 0, private.endptr);
>  
>          /* parse ENDSEG if passed */
> @@ -933,7 +939,8 @@ main(int argc, char **argv)
>          }
>  
>  
> -        if (!XLByteInSeg(private.endptr, segno) &&
> +        if (!XLogRecPtrIsInvalid(private.endptr) &&
> +            !XLByteInSeg(private.endptr, segno) &&
>              private.endptr != (segno + 1) * XLogSegSize)
>          {
>              fprintf(stderr,

Other than that it looks reasonable from a quick glance.

Andres

On Mon, Jul 18, 2016 at 8:10 PM, Magnus Hagander <magnus@hagander.net> wrote:
>
>
> On Thu, Jul 14, 2016 at 8:27 PM, Andres Freund <andres@anarazel.de> wrote:
>>
>> On 2016-07-14 13:46:23 +0200, Magnus Hagander wrote:
>> > Currently, if you run pg_xlogdump with -f, you have to specify an end
>> > position in an existing file, or if you don't it will only follow until
>> > the
>> > end of the current file.
>>
>> That's because specifying a file explicitly says that you only want to
>> look at that file, specifying two files that you want the range
>> inclusively between the two files.  -f works if you just use -s.
>
>
> Hmm. It does now. I'm *sure* it didn't when I was testing it. It must've
> been something else that was broken at that point :)

Same as Andres here, my understanding is that one file means that you
just want to look at this file, and two files permits to look at a
range of changes. But I have no argument against changing the current
behavior either.

>> > I'd appreciate a review of that by someone who's done more work on the
>> > xlog
>> > stuff, but it seems trivial to me. Not sure I can argue it's a bugfix
>> > though, since the usecase simply did not work...
>>
>> I'd say it's working as intended, and you want to change that
>> intent. That's fair, but I'd not call it a bug, and I'd say it's not
>> really 9.6 material.
>
> Based on that, I agree that it's working as intended.
>
> And definitely that it's not 9.6 material.
>
> I'll stick it on the CF page so I don't forget about it.

Moved to next CF. Magnus, you may want to finish wrapping that if you
still intend to change the current behavior.
-- 
Michael

On Fri, Dec 2, 2016 at 1:36 PM, Haribabu Kommi <kommi.haribabu@gmail.com> wrote:
> Patch received feedback at the end of commitfest.
> Closed in 2016-11 commitfest with "moved to next CF".
> Please feel free to update the status once you submit the updated patch.

And the thread has died as well weeks ago. I am marking that as
"returned with feedback" as there are no new patches.
-- 
Michael