Re: [PATCH] XLogReader v2 - Mailing list pgsql-hackers

From Tom Lane
Subject Re: [PATCH] XLogReader v2
Date
Msg-id 26913.1347216038@sss.pgh.pa.us
Whole thread Raw
In response to Re: [PATCH] XLogReader v2  (Andres Freund <andres@2ndquadrant.com>)
Responses Re: [PATCH] XLogReader v2
List pgsql-hackers
Andres Freund <andres@2ndquadrant.com> writes:
> On Tuesday, September 04, 2012 09:33:54 PM Alvaro Herrera wrote:
>> * There are way too many #ifdef VERBOSE_DEBUG stuff for my taste.  It
>> might look better if you had macros such as elog_debug() that are defined
>> to empty if VERBOSE_DEBUG is not defined.  (The problem with such an
>> approach is that you have to get into the business of creating one macro
>> for each different param count, so elog_debug1(), elog_debug2() and so
>> on.  It also means you have to count the number of args in each call to
>> ensure you're calling the right one.)

> Hm. I am generally not very happy with the logging as is. I don't want to rely 
> on elog() at all because that means the code suddently depends on just about 
> the whole backend which sucks (see my god ulgy makefile hack for that...).

elog/ereport are already basically macros.  Can't they be redefined for
use in a standalone program, with just minimal backing code?

> If we were to use that approach is there a platform that stops us from using 
> vararg macros? I *think* it is C99...

C90 is still the project standard, and this is a pretty lame reason to
want to change it.

>> * In the code beautification front, there are a number of cuddled braces
>> and improperly indented function declarations.

> I never seem to get those right. I really tried to make a pass over the whole
> file correcting them...

Install pgindent?
        regards, tom lane



pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: build farm machine using mixed results
Next
From: Andrew Dunstan
Date:
Subject: Re: build farm machine using mixed results