Transliteration fixes:

  * Clear any substitution if codec reset has been requested.
  * Don't report memory exhaustion when failing to allocate space
    for the test conversion in translit_try_sequence: there's
    nothing the caller can do, so treat it as if the substitution
    cannot be converted to the target character set.
  * Correctly report success if we run out of input immediately
    following a flush of a substitution.

Additional tests for transliteration.

1 files changed, 13 insertions, 3 deletions

diff --git a/build/tools/gentranstab.pl b/build/tools/gentranstab.pl
index 0e9205a..1b1ccad 100644
--- a/build/tools/gentranstab.pl
+++ b/build/tools/gentranstab.pl
@@ -48,12 +48,18 @@ static int translit_try_sequence(struct encoding_context *e,
 	size_t orig_tmplen, tmplen, index;
 	int ret = 1;
 
-	/* First, determine if sequence can be written to target encoding */
+	/* Determine if sequence can be written to target encoding */
 	/* Worst case: conversion to UTF-8 (needing 6 bytes per character) */
 	orig_tmplen = tmplen = (seqlen + 1) * 6;
 	ptmpbuf = tmpbuf = malloc(tmplen);
-	if (tmpbuf == NULL)
-		return 0;
+	if (tmpbuf == NULL) {
+		/* Consider lack of memory an inability to write the output.
+		 * We cannot report memory exhaustion from here, as it will
+		 * result in the caller thinking that the output buffer is
+		 * too small, which isn't actually the case. As 
+		 * transliteration is best-effort anyway, this should be ok. */
+		return -1;
+	}
 
 	/* Reset the transout codec */
 	if (e->transout != NULL) {
@@ -102,6 +108,8 @@ int translit_flush_replacement(struct encoding_context *e)
 	size_t substlen = e->substlen;
 	int ret = 1;
 
+	LOG(("Flushing %zd characters", substlen));
+
 	while (substlen > 0) {
 		UCS4 c = substitution[0];
 		
@@ -118,6 +126,8 @@ int translit_flush_replacement(struct encoding_context *e)
 	e->substitution = substitution;
 	e->substlen = substlen;
 
+	LOG(("%zd characters remaining", substlen));
+
 	return ret;
 }