Re: lots of unused variable warnings in assert-free builds - Mailing list pgsql-hackers

From Tom Lane
Subject Re: lots of unused variable warnings in assert-free builds
Date
Msg-id 9212.1327428230@sss.pgh.pa.us
Whole thread Raw
In response to Re: lots of unused variable warnings in assert-free builds  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: lots of unused variable warnings in assert-free builds  (Robert Haas <robertmhaas@gmail.com>)
List pgsql-hackers
I wrote:
> Also, it occurs to me that an intermediate macro
> "PG_USED_FOR_ASSERTS_ONLY" would be a good idea, first because it
> documents *why* you want to mark the variable as possibly unused,
> and second because changing the macro definition would provide an easy way
> to check for totally-unused variables, in case we wanted to periodically
> make such checks.

Uh, wait a second.  Why not

#ifdef USE_ASSERT_CHECKING
#define PG_USED_FOR_ASSERTS_ONLY
#else
#define PG_USED_FOR_ASSERTS_ONLY __attribute__((unused))
#endif

Then, when you build with asserts on, you *automatically* get told
if the variable is entirely unused.
        regards, tom lane


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: lots of unused variable warnings in assert-free builds
Next
From: Robert Haas
Date:
Subject: Re: lots of unused variable warnings in assert-free builds