Thread

RE: libpq debug log

Aya Iwata (Fujitsu) <iwata.aya@fujitsu.com> — 2021-02-22T08:22:02Z

Hi Tsunakawa san,

I update patch to v19.

> -----Original Message-----
> From: Tsunakawa, Takayuki/綱川 貴之 <tsunakawa.takay@fujitsu.com>
> Sent: Monday, February 22, 2021 1:30 PM
> (52)
> +      of tracing.  If (<literal>flags</literal> contains
> <literal>PQTRACE_SUPPRESS_TIMESTAMPS</literal>),
> 
> () can be removed?

Yes, I removed ().

> (53)
> +	int			inLogging;		/* next byte of
> logging */
> 
> I understood this variable contains a position or offset from some place.  Then,
> isn't it better to call it LogPos, LogOffset, or LogCursor?  I don't think you
> need to prepend "In" as in InCursor, because you don't have to distinguish
> between input and output unlike InCursor in PGconn.

Yes, "In" is from inCursor. I change inLogging to LogCursor because this name easier to understand 
the meaning of this parameter.

> (54)
> Can't you restrict the use of the above InLogging in fe-logging.c?  How about
> making InLogging mean the offset from InStart, i.e. the offset from the start of
> a message?

I fixed it.
Remove inLogging code from pqCheckInBufferSpace() and pqReadData().
And LogCursor remember the difference between conn->inCursor and conn->inStart.

And 3 bugs I noted February 2nd email are all fixed.
> 1. fix 3 bugs
>   1.1 -1 output in "Query" message
>   1.2 two message output in "ReadyForQuery" message
>   1.3 "StartupMessage" output as " UnknownMessage "

If you want to know how to fix there bugs, please see my email;
> Sent: Monday, February 8, 2021 11:58 PM

Regards,
Aya Iwata
Fujitsu