conchuela doesn't like gnu_printf anymore - Mailing list pgsql-hackers

From Tom Lane
Subject conchuela doesn't like gnu_printf anymore
Date
Msg-id 120507.1659805164@sss.pgh.pa.us
Whole thread Raw
Responses Re: conchuela doesn't like gnu_printf anymore
List pgsql-hackers
Buildfarm animal conchuela recently started spitting a lot of warnings
like this one:

 conchuela     | 2022-08-06 12:35:46 | /home/pgbf/buildroot/HEAD/pgsql.build/../pgsql/src/include/port.h:208:70:
warning:'format' attribute argument not supported: gnu_printf [-Wignored-attributes] 

I first thought we'd broken something, but upon digging through the
buildfarm history, the oldest build showing these warnings is

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=conchuela&dt=2022-07-18%2020%3A20%3A18

The new commits in that build don't look related, but what does look
related is that the choice of C++ compiler changed:

configure: using compiler=gcc 8.3 [DragonFly] Release/2019-02-22
configure: using CXX=ccache clang++14

vs

configure: using compiler=gcc 8.3 [DragonFly] Release/2019-02-22
configure: using CXX=g++

This is seemingly an intentional configuration change, because the
animal is reporting different config_env than before.  However,
we decide what to set PG_PRINTF_ATTRIBUTE to based on what CC
likes, and if CXX doesn't like it then you'll get these warnings.
(The warnings only appear in C++ compiles, else there'd REALLY
be a lot of them.)

Is it worth the trouble to try to set PG_PRINTF_ATTRIBUTE differently
in C and C++ builds?  I doubt it.  Probably the right fix for this
is to use matched C and C++ compilers, either both clang or both gcc.
I fear that inconsistency could lead to bigger problems than some
warnings.

            regards, tom lane



pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: A cost issue in ORDER BY + LIMIT
Next
From: Nathan Bossart
Date:
Subject: Re: optimize lookups in snapshot [sub]xip arrays