[JulienSchmidt]

Bug #3 (The race) was silently introduced by a semantic change in Go's database/sql in January 2018: https://github.com/golang/go/commit/651ddbdb5056ded455f47f9c...

It took until December of the same year until we got the first bug report and figured out what was going on. While the semantic change might look subtle, it was certainly not from our (driver maintainer's) perspective. We were quite disappointed that such a change was made 1) without informing the driver maintainers 2) making sure the changes were in place before this change made it into a Go release.

We regularly test against Go's master (now using a Travis CI cron job), but that only helps if the existing tests fail. We don't have the time to constantly monitor all changes in the Go repo.

If there is a need to make such changes (not just in database/sql and not just in Go), PLEASE actively communicate early with the community / the direct users.

[kardianos]

I'm sorry. This is clearly my fault and I'm sorry this wasn't communicated.

Please look into Go1.15. It should all be backwards compatible, but where the session resetter is called changes, and a new call is added for verifying if the connection is valid or not. I'll have the release notes updated soon for Go1.15, but until then, look at the driver docs for tip.

I'll make announce the change on the golang-sql mailing group.

[nemetroid]

A comment on the original issue[1] promises to ”[open] some tickets for various drivers”. Did this not happen?

1: https://github.com/golang/go/issues/23519#issuecomment-36057...

[JulienSchmidt]

I really appreciate the work that is done to improve the database/sql package. It is still rather young and not as mature as for example JDBC. I would rather thank the people who work on it for the work they do, than to point fingers for anything that is not done. The original design of database/sql was sometimes a bit too simplistic and will probably require some more ugly changes or workarounds in the future.

This specific change was a bit unfortunate. It tried to fix another bug caused by the introduction of Context cancellation support, but it unfortunately caused some other major bugs due to the changed semantics. Like the very same comment already indicates, other drivers also required changes. In retro-perspective I think it would have been better to try to handle the original issue entirely in database/sql, instead of changing the "contract" with the drivers.

Go MySQL Driver was, and I believe still is, developed mostly by some random programmers in their free time (I originally started it as a side-project during high school and until now was never paid for any work on it). It probably makes anyone who worked on it proud that now some major companies, not just GitHub, employ it. But like many such projects, it is not a perfect or finished product. If you use community-driven open source software at work, try to convince your management to set some work hours aside for actually contributing back to those, like GitHub did.

[networkimprov]

Go is still undergoing intensive R&D, especially in the runtime. Its releases are quietly tested against internal Google projects, but not formally tested with widely-used open source projects.

Any org that deploys Go apps should designate someone to watch the Go issue tracker, and subscribe to all issues that might affect them.

https://github.com/golang/go/issues/

[mfer]

Notifying users of the code of a semantic change does not mean it's not a semantic change.

Even if the Go team can find a way to notify all of the public drivers impacted, they aren't going to be able to know about the private drivers or forks to notify those people. Nor are they going to be able to notify all the consumers who, using MVS, have the version of the driver set to one that's not fixed.

Bugs from this change will propagate and there's nothing the Go team can do to stop them all. This is why projects following semantic versioning are not supposed to make breaking changes in minor releases.

This is, unfortunately, not the first time I've seen this kind of change in Go. Some previous changes have caused bugs for me a few times in the past. Not knowing which version of the compiler one would use I ended up needed to craft code to handle before and after cases for the change along with the bug reports from users.