Re: exposing pg_controldata and pg_config as functions

On Wed, Jan 20, 2016 at 2:32 AM, Joe Conway <mail(at)joeconway(dot)com> wrote:
> The only things I know of still lacking is:
> 1) Documentation
> 2) Decision on REVOKE ... FROM PUBLIC

Yep, regarding 2) I am the only one actually making noise to protect
this information by default, against at least 2 committers :)

> I'm assuming by the lack of complainants that there is enough support
> for the feature (conceptually at least) that it is worthwhile for me to
> write the docs. Will do that over the next couple of days or so.

I would think so as well.

+typedef struct configdata
+{
+ char *name;
+ char *setting;
+} configdata;
For a better analogy to ControlFileData, this could be renamed ConfigData?

-OBJS_COMMON = exec.o pg_lzcompress.o pgfnames.o psprintf.o relpath.o \
- rmtree.o string.o username.o wait_error.o
+# don't include subdirectory-path-dependent -I and -L switches
+STD_CPPFLAGS := $(filter-out -I$(top_srcdir)/src/include
-I$(top_builddir)/src/include,$(CPPFLAGS))
+STD_LDFLAGS := $(filter-out -L$(top_builddir)/src/port,$(LDFLAGS))
The point of the move to src/common is to remove the duplication in
src/bin/pg_config/Makefile.

--- /dev/null
+++ b/src/common/config_info.c
[...]
+#include "postgres.h"
+
+#include "miscadmin.h"
+#include "common/config_info.h"
All the files in src/common should begin their include declarations with that:
#ifndef FRONTEND
#include "postgres.h"
#else
#include "postgres_fe.h"
#endif

+configdata *
+get_configdata(char *my_exec_path, size_t *configdata_len)
+{
It may be good to mention that the result is palloc'd and that caller
may need to free it if necessary. It does not matter in the two code
paths of this patch, but it may matter for other users calling that.

MSVC compilation is working correctly here.
--
Michael