Correct possible dropping of notnull from locale columns on upgrade

[asmecher]

The implementation of #9425 included migration code that extended the length of any identified locale columns. However, some of these columns (e.g. publication_galleys.locale) are nullable, and this code will either make them NOT NULL or, if there is null data present, fail with an error.

I would suggest altering the migration to make it preserve the NOT NULL status of the column when extending the length.

@jyhein and @ajnyga, just a heads-up -- @touhidurabir would like to fix it.

PRs
pkp-lib --> #10305
ojs --> pkp/ojs#4398 [TEST ONLY]

[ajnyga]

thanks, go ahead

[touhidurabir]

@asmecher tests are passing , can you take a look at the PR at #10293 (comment)

[asmecher]

@ajnyga, could you or @jyhein review this one? Thanks!

[asmecher]

@ajnyga or @jyhein, this is breaking just about everything on upgraded installations; could you have a look at the PRs above?

[bozana]

@touhidurabir, what does "->first(default:[])" mean?

[touhidurabir]

@touhidurabir, what does "->first(default:[])" mean?

@bozana please see at https://github.com/pkp/pkp-lib/pull/10305/files#r1771286012 for reasoning .

[bozana]

So I took a look at the code changes and they look good to me: we now look if the column to be changed is nullable and act accordingly i.e. we use $table->string($l, 28)->nullable()->change() (instead of $table->string($l, 28)->change()) in that case.
I have not tested it, but I think we can merge it :-)

[ajnyga]

sorry for missing this in august, we talked about this in the morning and looks to be fine. I also tried running upgrade test from 3.4 stable.

[asmecher]

Merged your PR, thanks, @touhidurabir!