Fix incompatibility with bootsnap

[dsander]

When using bootsnap in production it tries to compile the migration
files, but fails because the files are actually erb and thus not a
valid ruby file.

[albus522]

What specific error did you run into and under what configuration? I have not seen where bootstrap is trying to compile these files.

[dsander]

The error was that files were named .rb while actually being .erb templates and not containing valid ruby syntax.

[albus522]

I was asking for the actual error message as I can not reproduce your issue. Those files are never required by ruby and therefore should never be touched by bootsnap. A backtrace from the error would also be helpful.

[dsander]

Sorry you are right, at Huginn we have been running on a fork that uses this commit: dsander@55535e2 I kind of forgot about that change and without it (and the latest master branch) it works without issues.

Kind of unrelated question, we conditionally need to modify the reserve_sql_strategy, but since the Railtie is running after_initialize we can not do that in a Rails initializer.

Would you accept a PR that adds a ActiveSupport.run_load_hooks call like this: dsander@0b433cc ?

Thanks for your recent work on this gem!

[albus522]

I am reasonable sure after_initialize hooks are run in the order in which they are defined. So if you add an after_initialize in your app that makes the changes you want, it should run after the one defined in DJ runs.

[albus522]

Thank you for pointing out the reserve_sql_strategy issue. I need to update the documentation about that. I knew that change had something we were missing.

[dsander]

it should run after the one defined in DJ runs.

Thanks that seems to work. 👍

[albus522]

@dsander I have been picking away at #149 slowly and would be curious to know what caused you to choose the old lock format.

[dsander]

We are only using the old lock strategy on mysql. Here is the original issue and reproduction: huginn/huginn#977 (comment). I believe we tested it against Postgres as well and did not have any issues with the optimized strategy.

[albus522]

Thank you for digging that up.