Re: speed up unicode decomposition and recomposition

[Michael Paquier]

On Fri, Oct 23, 2020 at 04:18:13PM -0700, Mark Dilger wrote:
> On Oct 23, 2020, at 9:07 AM, Tom Lane <t...@sss.pgh.pa.us> wrote:
>> genhtml: WARNING: function data mismatch at 
>> /home/postgres/pgsql/src/common/unicode_norm.c:102
>> 
>> I've never seen anything like that before.  I suppose it means that
>> something about 783f0cc64 is a bit fishy, but I don't know what.
>> 
>> The emitted coverage report looks fairly normal anyway.  It says
>> unicode_norm.c has zero test coverage, which is very possibly correct
>> since I wasn't running in UTF8 encoding, but I'm not entirely sure of
>> that either.
> 
> I don't see it on mac nor on ubuntu64.  I get 70.6% coverage of
> lines and 90.9% of functions on ubuntu.

I can see the problem on Debian GID with lcov 1.14-2.  This points to
the second declaration of get_code_entry().  I think that genhtml,
because it considers the code of unicode_norm.c as a whole without its
CFLAGS, gets confused because it finds the same function to index as
defined twice.  It expects only one definition, hence the warning.  So
I think that this can lead to some incorrect data in the HTML report,
and the attached patch takes care of fixing that.  Tom, does it take
care of the issue on your side?
--
Michael

diff --git a/src/common/unicode_norm.c b/src/common/unicode_norm.c
index 4ffce0e619..7cc8faa63a 100644
--- a/src/common/unicode_norm.c
+++ b/src/common/unicode_norm.c
@@ -53,11 +53,26 @@
  * The backend version of this code uses a perfect hash function for the
  * lookup, while the frontend version uses a binary search.
  */
-#ifndef FRONTEND
+#ifdef FRONTEND
+/* comparison routine for bsearch() of decomposition lookup table. */
+static int
+conv_compare(const void *p1, const void *p2)
+{
+	uint32		v1,
+				v2;
+
+	v1 = *(const uint32 *) p1;
+	v2 = ((const pg_unicode_decomposition *) p2)->codepoint;
+	return (v1 > v2) ? 1 : ((v1 == v2) ? 0 : -1);
+}
+
+#endif
 
 static const pg_unicode_decomposition *
 get_code_entry(pg_wchar code)
 {
+#ifndef FRONTEND
+
 	int			h;
 	uint32		hashkey;
 	pg_unicode_decompinfo decompinfo = UnicodeDecompInfo;
@@ -82,33 +97,17 @@ get_code_entry(pg_wchar code)
 
 	/* Success! */
 	return &decompinfo.decomps[h];
-}
 
 #else
 
-/* comparison routine for bsearch() of decomposition lookup table. */
-static int
-conv_compare(const void *p1, const void *p2)
-{
-	uint32		v1,
-				v2;
-
-	v1 = *(const uint32 *) p1;
-	v2 = ((const pg_unicode_decomposition *) p2)->codepoint;
-	return (v1 > v2) ? 1 : ((v1 == v2) ? 0 : -1);
-}
-
-static const pg_unicode_decomposition *
-get_code_entry(pg_wchar code)
-{
 	return bsearch(&(code),
 				   UnicodeDecompMain,
 				   lengthof(UnicodeDecompMain),
 				   sizeof(pg_unicode_decomposition),
 				   conv_compare);
+#endif
 }
 
-#endif							/* !FRONTEND */
 
 /*
  * Given a decomposition entry looked up earlier, get the decomposed

signature.asc
Description: PGP signature