The bug csv01.test hangs on powerpc32 and powerpc64
(both are 'char' == 'unsigned char' platforms)
Hangup happens in
static int csvtabNext(sqlite3_vtab_cursor *cur){
...
if( z==0 || pCur->rdr.cTerm==EOF ){
Here 'pCur->rdr.cTerm' is of type 'char' (holds truncated EOF)
while EOF constant is (int)(-1).
On 'signed char' platforms both arguments sign-extend to 'int'
before comparison and test "works". But ont on 'unsigned char'
platforms where comparison is '0xff == -1'.
Workaround by trunacting EOF constant down to 'char' size.
Not an ideal fix but makes EOF handling more consistent.
Reported-by: Matt Turner
Bug: https://bugs.gentoo.org/630698
Signed-off-by: Sergei Trofimovich <sly...@gentoo.org>
---
ext/misc/csv.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/ext/misc/csv.c b/ext/misc/csv.c
index 6d99634..1bd9888 100644
--- a/ext/misc/csv.c
+++ b/ext/misc/csv.c
@@ -686,7 +686,13 @@ static int csvtabNext(sqlite3_vtab_cursor *cur){
pCur->aLen[i] = 0;
i++;
}
- if( z==0 || pCur->rdr.cTerm==EOF ){
+ /*
+ * HACK: truncate EOF to 'char' to make code
+ * work equally (bad) on systems with signed
+ * and unsigned char. Other wise platforms
+ * with unsigned char loop indefinitely here.
+ */
+ if( z==0 || pCur->rdr.cTerm==(char)EOF ){
pCur->iRowid = -1;
}else{
pCur->iRowid++;
--
2.14.1
_______________________________________________
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
