Re: pgsql: Mark the second argument of pg_log as the translatable string in

On Tue, Apr 14, 2015 at 2:17 AM, Alvaro Herrera
<alvherre(at)2ndquadrant(dot)com> wrote:
> Michael Paquier wrote:
>> On Sun, Apr 12, 2015 at 10:17 AM, Alvaro Herrera wrote:
>> > What pg_basebackup's progress_report() does is have the message in the
>> > translatable part not include the \r; the \r is in a separate fprintf()
>> > call.
>>
>> Like the attached then.
>
> Not a fan of this approach, because now this function knows that
> pg_log(PG_PROGRESS) is equivalent to printf(). This abstraction is a
> bit leaky, isn't it ... Probably not worth sweating about, though.
>
>> diff --git a/src/bin/pg_rewind/logging.c b/src/bin/pg_rewind/logging.c
>> index aba12d8..3e2dc76 100644
>> --- a/src/bin/pg_rewind/logging.c
>> +++ b/src/bin/pg_rewind/logging.c
>> @@ -134,7 +134,8 @@ progress_report(bool force)
>> snprintf(fetch_size_str, sizeof(fetch_size_str), INT64_FORMAT,
>> fetch_size / 1024);
>>
>> - pg_log(PG_PROGRESS, "%*s/%s kB (%d%%) copied\r",
>> + pg_log(PG_PROGRESS, "%*s/%s kB (%d%%) copied",
>> (int) strlen(fetch_size_str), fetch_done_str, fetch_size_str,
>> percent);
>> + printf("\r");
>> }

So could you elaborate your "favorite" approach?

Now pg_log() calls printf() and fflush(stdout). So '\r' is printed after fflush.
It's a bit strange. Maybe we can just replace pg_log() with printf() here.

Another question is; should we output the progress report to stderr rather
than stdout? I thought this because I found that pg_basebackup reports
the progress to stderr.

Regards,

--
Fujii Masao