Remove unnecessary type violation in tsvectorrecv().

compareentry() is declared to work on WordEntryIN structs, but
tsvectorrecv() is using it in two places to work on WordEntry
structs.  This is almost okay, since WordEntry is the first
field of WordEntryIN.  But on machines with 8-byte pointers,
WordEntryIN will have a larger alignment spec than WordEntry,
and it's at least theoretically possible that the compiler
could generate code that depends on the larger alignment.

Given the lack of field reports, this may be just a hypothetical bug
that upsets nothing except sanitizer tools.  Or it may be real on
certain hardware but nobody's tried to use tsvectorrecv() on such
hardware.  In any case we should fix it, and the fix is trivial:
just change compareentry() so that it works on WordEntry without any
mention of WordEntryIN.  We can also get rid of the quite-useless
intermediate function WordEntryCMP.

Bug: #18875
Reported-by: Alexander Lakhin <[email protected]>
Author: Tom Lane <[email protected]>
Discussion: https://postgr.es/m/18875-07a29c49c825a608@postgresql.org
Backpatch-through: 13

diff --git a/src/backend/utils/adt/tsvector.c b/src/backend/utils/adt/tsvector.c
index 650be842f28febbd7c54d17955895bc3e4f108db..1fa2e3729bfab6391f61fe8a1279dd32f5a69656 100644 (file)
--- a/src/backend/utils/adt/tsvector.c
+++ b/src/backend/utils/adt/tsvector.c
@@ -25,7 +25,7 @@
 
 typedef struct
 {
-   WordEntry   entry;          /* must be first! */
+   WordEntry   entry;          /* must be first, see compareentry */
    WordEntryPos *pos;
    int         poslen;         /* number of elements in pos */
 } WordEntryIN;
@@ -79,16 +79,19 @@ uniquePos(WordEntryPos *a, int l)
    return res + 1 - a;
 }
 
-/* Compare two WordEntryIN values for qsort */
+/*
+ * Compare two WordEntry structs for qsort_arg.  This can also be used on
+ * WordEntryIN structs, since those have WordEntry as their first field.
+ */
 static int
 compareentry(const void *va, const void *vb, void *arg)
 {
-   const WordEntryIN *a = (const WordEntryIN *) va;
-   const WordEntryIN *b = (const WordEntryIN *) vb;
+   const WordEntry *a = (const WordEntry *) va;
+   const WordEntry *b = (const WordEntry *) vb;
    char       *BufferStr = (char *) arg;
 
-   return tsCompareString(&BufferStr[a->entry.pos], a->entry.len,
-                          &BufferStr[b->entry.pos], b->entry.len,
+   return tsCompareString(&BufferStr[a->pos], a->len,
+                          &BufferStr[b->pos], b->len,
                           false);
 }
 
@@ -167,12 +170,6 @@ uniqueentry(WordEntryIN *a, int l, char *buf, int *outbuflen)
    return res + 1 - a;
 }
 
-static int
-WordEntryCMP(WordEntry *a, WordEntry *b, char *buf)
-{
-   return compareentry(a, b, buf);
-}
-
 
 Datum
 tsvectorin(PG_FUNCTION_ARGS)
@@ -511,7 +508,7 @@ tsvectorrecv(PG_FUNCTION_ARGS)
 
        datalen += lex_len;
 
-       if (i > 0 && WordEntryCMP(&vec->entries[i],
+       if (i > 0 && compareentry(&vec->entries[i],
                                  &vec->entries[i - 1],
                                  STRPTR(vec)) <= 0)
            needSort = true;