Bug - insert statements fail under yet to be determined conditions

[LeGEC]

I stumbled on a bug involoving the interaction between the driver and MySQL.

After much trial & error, I managed to extract the following test cases :
gist : extra_mysql_test.go

This file contains 5 test cases.
On my machine : Tests 1-2-3 pass, tests 4-5 fail. I would expect the 5 tests to pass.

I would like to know if it fails on other configurations.

Go version : go1.2.1 linux/amd64
MySQL version : the original bug was triggered under MySQL 5.1.63 (Debian squeeze), and MySQL 5.5.35 (Ubuntu 13.10).

[LeGEC]

Extra details about the tests :

All cases involve the following :

• inserting multiple rows,
• in a table which contains a nullable column,
• using, on the go side, sql.NullInt64 values.

What changes between the tests :

• the number of columns in the table (TEST 1 table has 7 columns, the others have 8)
• the number of rows inserted at once (TEST 2 inserts 9 rows, the others insert 10 rows)
• (consequence of the two previous points :) the number of values passed along with a statement
• some variation on what is inserted in the nullable column (null values versus non null values)

I was not able to find the exact source of the bug, but there seems to be a threshold on the amount of sent parameters, over which these insert statements start to fail.

The "tipping point" which I found is between "TEST 3" and "TEST 4" :

• both of these tests execute a 7-columns insert statement,
• both of these tests insert 10 rows together,
• the only difference lies in the value inserted in the last row (9 vs NULL)

I will try to look for more details, but I am not yet familiar with the internals of the driver.

[arnehormann]

I'll dig deeper into this later, but first I'd like you to comment on a hunch of mine.
Does the driver version you use include #210?
The referenced pull request also includes a test for prepared statements.

[arnehormann]

@LeGEC I checked this with the current master - all your tests already pass. You are apparently using an older version of the driver. That's pretty easy - the current master is not go getable yet, the pull request is not part of a released version.
Thanks for hunting this down, the next release will contain the fix.
If you update to the most recent version and this still fails, please add a comment and I will reopen this issue.

[LeGEC]

@arnehormann The tests do indeed pass on master. Thx for the pointer. Do you have a planned release date ?

[julienschmidt]

No fixed date yet, but "soon": https://github.com/go-sql-driver/mysql/issues?milestone=6&state=open

[LeGEC]

@arnehormann, @julienschmidt : How can I make this fix "go gettable" ?

I forked the project, how should I set up the tags on the fork so that "go get --fork--/mysql" gets me the fix ?

Thanks in advance

[julienschmidt]

git tag -f go1, then push the changed tag to the repo with the --tags option

[LeGEC]

Thanks, worked like a charm.

btw, the fork is :

https://github.com/LeGEC/mysql