Restart app on pom.xml change

[stuartwdouglas]

Unlike other hot replacement this simply watches the pom.xml files,
and reloads the complete application on change. This means there is
a short period where the app in unavailible.

Fixes #4871

[jaikiran]

Hello @stuartwdouglas, is this meant to be passed along to the new DevModeRunner, so that it doesn't (start and) suspend in debug mode? If so, I don't see changing this value playing a role in it, since the args being passed to the new DevModeRunner already have a bunch of JVM -Xdebug options setup in it.

[stuartwdouglas]

Good point, I will just move all this logic into prepare()

[emmanuelbernard]

@stuartwdouglas I think it works well I tested it in a few combinations and never had unexpected behaviors. Except in one case:

• create a new RESTEasy project (from generator)
• add mvn quarkus:add-extensions -Dextensions="jdbc-postgresql,hibernate-orm-panache"
• add Todo class with @entity on it
• refresh see error
• remove jdbc-postgresql,hibernate-orm-panache fromt he pom
• refresh and don't see error even though the Todo app has missing packages and annotation references that would not compile

We can survive without but that's a bit of an unexpected bahavior.

Also needs testing on Windows I think @maxandersen

[emmanuelbernard]

Stuart I approved it with two pending conditions:

• review my "non compilation" error comment and see if that's an easy fix
• I asked @maxandersen to test this PR on Windows, it's probably smart to make sure it works on W before pushing into the release. He will od it today asap after one platform issue is behind us

[emmanuelbernard]

@stuartwdouglas ^

[stuartwdouglas]

There is nothing I can do about that short of issuing a 'clean compile' invocation after pom changes. Even then I am not really sure how I would go about reporting the error.

I don't think this is a big deal as mostly we care about adding extensions rather than removing them.

[maxandersen]

first thing - if you clean compile don't you risk loosing any post-compile generated artifacts in the target dir ? or are they assumed not relevant for devmode ?

[stuartwdouglas]

dev mode running happens straight after compile, so if there are any they will be out of date anyway. I don't think it is really a workable solution, and I don't think it is a major problem.

[emmanuelbernard]

Yes I'm cool with that little quirk and see how it develops in practical usages

[maxandersen]

for the record i tested the basic flow and worked on windows.