I read pg_upgrade code glance over, and found 4 issues in it.
Are there any issues to be fixed before 9.0 release?
1. NAMEDATASIZE 2. extern PGDLLIMPORT 3. pathSeparator 4. EDB_NATIVE_LANG
==== 1. NAMEDATASIZE ====
pg_upgrade has the following definition, but should it be just NAMEDATALEN?
/* Allocate for null byte */ #define NAMEDATASIZE (NAMEDATALEN + 1)
Table names should be in NAMEDATELEN - 1 bytes. At least 64th bytes in
"name" data is always '\0'.
=# CREATE TABLE "1234567890...(total 70 chars)...1234567890" (i int); NOTICE: identifier "123...890" will be
truncatedto "123...0123"
==== 2. extern PGDLLIMPORT ====
pg_upgrade has own definitions of extern PGDLLIMPORT Oid binary_upgrade_next_xxx
in pg_upgrade_sysoids.c. But those variables are not declared as
PGDLLIMPORT in the core. Can we access unexported variables here?
==== 3. pathSeparator ====
Path separator for Windows is not only \ but also /. The current code
ignores /. Also, it might not work if the path string including multi-byte
characters that have \ (0x5c) in the second byte.
==== 4. EDB_NATIVE_LANG ====
Of course it is commented out with #ifdef, but do we have codes
for EDB in core?
Regards,
---
Takahiro Itagaki
NTT Open Source Software Center