don't be strict in test for warnings

[arnehormann]

No description provided.

[arnehormann]

It's the first one that came to mind - I think every single one would be confusing as there's no "RELAXED_DATA_LENGTH" setting afaik. At least we use this one somewhere else...
What does "Yes" mean? Yes add the parameter or Yes make a quickfix (I read it as make and document the quickfix).

[julienschmidt]

Yes to the parameter, sorry.

[julienschmidt]

What about TRADITIONAL mode?
http://dev.mysql.com/doc/refman/5.0/en/server-sql-mode.html#sqlmode_traditional

Edit: Ignore this comment. It does the opposite of what you intended.

[arnehormann]

Just tested it: for me, the test still fails in traditional mode

[arnehormann]

So are we also introducing PTAL after LGTM? Implemented the dsn parameter - but it wasn't necessary to add it to DBTest. I like the way it looks now, no "quickfix" comment required.
... PTAL 😃

[julienschmidt]

LGTM, SGTM, PTAL ... everything you want ;-)

I'm still wondering why the invalid dates mode works... The documentation says nothing about other behaviour changes than allowing invalid date ranges.
http://dev.mysql.com/doc/refman/5.0/en/server-sql-mode.html#sqlmode_allow_invalid_dates

@xaprb ? :)

[julienschmidt]

But LGTM for now as a hotfix. We still can swap the sql mode later.

[arnehormann]

IMO, the invalid dates mode works because it sets the SQL mode to something not including STRICT_TABLES or STRICT_TRANS_TABLES - which convert overflow warnigns to errors. TRADITIONAL includes STRICT_TRANS_TABLES. ALLOW_INVALID_DATES is the only "relaxed" mode I found which doesn't introduce syntax changes.
I tried to set it to "nothing" after the = - but that didn't work either.

[arnehormann]

Btw., my config if someone wants to reproduce it:

I' using MySQL 5.6.10 (on OS X, installed via homebrew).

From my.cnf
sql_mode=STRICT_TRANS_TABLES,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_AUTO_VALUE_ON_ZERO,NO_ENGINE_SUBSTITUTION,NO_ZERO_DATE,NO_ZERO_IN_DATE,ONLY_FULL_GROUP_BY