Pkg.test() is failing on Windows 64 bit #40
Comments
|
I also can't run the tests on x64 windows with a similar error at the same place. It looks like NAs are not being returned correctly. I also had to fiddle with The error is in reading the results - the values appear in MySQL correctly (and the results-as-tuples test works):
|
|
A bit more isolation: it's specifically related to bound select statements and not to either tuples or dataframes:
Running the second cell repeatedly produced different types of garbage in place of the nulls each time. As best I can tell, the prepared statement version calls |
|
The MySQL bug that I mentioned in the comment in mysql_execute(hndl, "DROP USER IF EXISTS 'test'@'127.0.0.1';" )The NA's not showing up is caused by the fact that |
|
It looks like I made a mistake while defining the |
|
I tried to make this change but I don't actually see where there should be a typedef struct st_mysql_bind
{
unsigned long *length; /* output length pointer */
my_bool *is_null; /* Pointer to null indicator */
void *buffer; /* buffer to get/put data */
/* set this if you want to track data truncations happened during fetch */
my_bool *error;
unsigned char *row_ptr; /* for the current data position */
void (*store_param_func)(NET *net, struct st_mysql_bind *param);
void (*fetch_result)(struct st_mysql_bind *, MYSQL_FIELD *,
unsigned char **row);
void (*skip_result)(struct st_mysql_bind *, MYSQL_FIELD *,
unsigned char **row);
/* output buffer length, must be set when fetching str/binary */
unsigned long buffer_length;
unsigned long offset; /* offset position for char/binary fetch */
unsigned long length_value; /* Used if length is 0 */
unsigned int param_number; /* For null count and error messages */
unsigned int pack_length; /* Internal length for packed data */
enum enum_field_types buffer_type; /* buffer type */
my_bool error_value; /* used if error is 0 */
my_bool is_unsigned; /* set if integer type is unsigned */
my_bool long_data_used; /* If used with mysql_send_long_data */
my_bool is_null_value; /* Used if is_null is 0 */
void *extension;
} MYSQL_BIND;and I don't see a mysql_bind.h anywhere. I think I see two fields in MYSQL_STMT that should be immutable MYSQL_STMT # This is different in mariadb header file.
mem_root::MEM_ROOT
list::LIST
mysql::Ptr{Void}
params::MYSQL_BIND
bind::MYSQL_BIND
fields::MYSQL_FIELD
result::MYSQL_DATA
data_cursor::MYSQL_ROWS
affected_rows::Culonglong #Changed
insert_id::Culonglong #Changed
stmt_id::Culong
flags::Culong
prefetch_rows::Culong
server_status::Cuint
last_errno::Cuint
param_count::Cuint
field_count::Cuint
state::Cuint
last_error::Ptr{Cchar}
sqlstate::Ptr{Cchar}
send_types_to_server::Cint
bind_param_done::Cint
bind_result_done::Cuchar
unbuffered_fetch_cancelled::Cint
update_max_length::Cint
extension::Ptr{Cuchar}
endFor the offsets I see immutable MYSQL_BIND
length::Ptr{Culong} # 8
is_null::Ptr{Cchar} # 8
buffer::Ptr{Void} # 8
error::Ptr{Cchar} # 8
row_ptr::Ptr{Cuchar} # 8
store_param_func ::Ptr{Void} # 8
fetch_result ::Ptr{Void} # 8
skip_result ::Ptr{Void} # 8
buffer_length::Culong # sizeof(Clong)
offset::Culong # sizeof(Clong)
length_value::Culong # sizeof(Clong)
param_number :: Cuint # 4
pack_length :: Cuint # 4
buffer_type :: Cint # 4
error_value :: Cchar # 1
is_unsigned :: Cchar # 1
long_data_used :: Cchar # 1
is_null_value :: Cchar
extension :: Ptr{Void}so it looks like the offset should be Miraculously tests pass for me after this change, but someone should check and make sure I didn't break it for other systems. I'll submit a pull request mostly to see if it passes travis. |
Pkg.test() is failing while executing prepared statements when the value is NULL, I've included the exact error message below
Configuration
The text was updated successfully, but these errors were encountered: