Re: Proposal: knowing detail of config files via SQL

On 2/27/15 11:27 PM, Stephen Frost wrote:
>> >@@ -344,6 +346,21 @@ ProcessConfigFile(GucContext context)
>> >                          PGC_BACKEND, PGC_S_DYNAMIC_DEFAULT);
>> >      }
>> >
>> >+    guc_file_variables = (ConfigFileVariable *)
>> >+            guc_malloc(FATAL, num_guc_file_variables * sizeof(struct ConfigFileVariable));
> Uh, perhaps I missed it, but what happens on a reload?  Aren't we going
> to realloc this every time?  Seems like we should be doing a
> guc_malloc() the first time through but doing guc_realloc() after, or
> we'll leak memory on every reload.
>
>> >+    /*
>> >+     * Apply guc config parameters to guc_file_variable
>> >+     */
>> >+    guc = guc_file_variables;
>> >+    for (item = head; item; item = item->next, guc++)
>> >+    {
>> >+        guc->name = guc_strdup(FATAL, item->name);
>> >+        guc->value = guc_strdup(FATAL, item->value);
>> >+        guc->filename = guc_strdup(FATAL, item->filename);
>> >+        guc->sourceline = item->sourceline;
>> >+    }
> Uh, ditto and double-down here.  I don't see a great solution other than
> looping through the previous array and free'ing each of these, since we
> can't depend on the memory context machinery being up and ready at this
> point, as I recall.

MemoryContextInit() happens near the top of main(), before we call 
InitializeGUCOptions(). So it should be possible to use memory contexts 
here. I don't know why guc doesn't use palloc; perhaps for historical 
reasons at this point?
-- 
Jim Nasby, Data Architect, Blue Treble Consulting
Data in Trouble? Get it in Treble! http://BlueTreble.com