Re: Unused header file inclusion

Andres Freund <andres@anarazel.de>

From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, vignesh C <vignesh21@gmail.com>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2019-08-19T20:07:50Z
Lists: pgsql-hackers
Hi,

On 2019-08-18 14:37:34 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > I've pushed the other ones.
> 
> Checking whether header files compile standalone shows you were overly
> aggressive about removing fmgr.h includes:
> 
> In file included from /tmp/headerscheck.Ss8bVx/test.c:3:
> ./src/include/utils/selfuncs.h:143: error: expected declaration specifiers or '...' before 'FmgrInfo'
> ./src/include/utils/selfuncs.h:146: error: expected declaration specifiers or '...' before 'FmgrInfo'
> ./src/include/utils/selfuncs.h:152: error: expected declaration specifiers or '...' before 'FmgrInfo'

Darn.  Pushed the obvious fix of adding a direct fmgr.h include, rather
than the preivous indirect include.


> That's with a script I use that's like cpluspluscheck except it tests
> with plain gcc not g++.  I attached it for the archives' sake.
> 
> Oddly, cpluspluscheck does not complain about those cases, but it
> does complain about

Hm. I don't understand why it's not complaining. Wonder if it's a
question of the flags or such.


> In file included from /tmp/cpluspluscheck.FgX2SW/test.cpp:4:
> ./src/bin/scripts/scripts_parallel.h:18: error: ISO C++ forbids declaration of 'PGconn' with no type
> ./src/bin/scripts/scripts_parallel.h:18: error: expected ';' before '*' token
> ./src/bin/scripts/scripts_parallel.h:29: error: 'PGconn' has not been declared

I noticed that "manually" earlier, when looking at the openbsd issue.


> (My headerscheck script is missing that header; I need to update it to
> match the latest version of cpluspluscheck.)

I wonder if we should just add a #ifndef HEADERCHECK or such to the
headers that we don't want to process as standalone headers (or #ifndef
NOT_STANDALONE or whatever). That way multiple tools can rely on these markers,
rather than copying knowledge about that kind of information into
multiple places.

I wish we could move the whole logic of those scripts into makefiles, so
we could employ parallelism.

Greetings,

Andres Freund



Commits

  1. Add fmgr.h include to selfuncs.h.

  2. Add "headerscheck" script to test header-file compilability under C.

  3. Fix failure-to-compile-standalone in scripts_parallel.h.

  4. Remove fmgr.h includes from headers that don't really need it.

  5. Don't include utils/array.h from acl.h.

  6. Remove redundant prototypes for SQL callable functions.

  7. Remove unnecessary #include <limits.h>

  8. Teach tuplesort.c about "top N" sorting, in which only the first N tuples