RFR: 8308588: Unnecessary synchronized on GTKStyle#ICONS_MAP can be removed [v2]

[Alexey Ivanov]

On Thu, 29 Aug 2024 12:42:49 GMT, Alexey Ivanov <aivanov at openjdk.org> wrote:

>>> > One slight concern is that, does GTKStockIcon has any role to play with synchronized block?
>>> 
>>> Sorry, I didn't understand what do you mean.
>> 
>> I meant, since `ICONS_MAP` stores `GTKStockIcon`  as value, I was wondering does that make any difference or need for the map to be synchronized? I also came across Immutable Maps which are inherently thread safe. If ICONS_MAP are immutable one, then Synchronized block is definitely redundant.
>
>> > > One slight concern is that, does GTKStockIcon has any role to play with synchronized block?
>> > 
>> > 
>> > Sorry, I didn't understand what do you mean.
>> 
>> I meant, since `ICONS_MAP` stores `GTKStockIcon` as value, I was wondering does that make any difference or need for the map to be synchronized?
> 
> In fact, `synchronized` block around `ICONS_MAP.get` had no effect.
> 
> To be correct, the value for `ICONS_MAP` had to be assigned inside a `synchronized` block too. Otherwise, it is not thread-safe.
> 
>> I also came across Immutable Maps which are inherently thread safe. If ICONS_MAP are immutable one, then Synchronized block is definitely redundant.
> 
> Effectively, `ICONS_MAP` is an immutable map. It's initialised in a static initialiser and is never modified afterwards.
> 
> As far as I can tell, `GTKStockIcon` objects stored in the map aren't modified either.
> 
> Thus, the entire structure is immutable and is thread-safe without additional synchronisation.

> @aivanov-jdk I guess `Maps.of` can't be used here because there are few elements put inside a map after `tk.checkGtkVersion(3, 10, 0)` condition…

It's possible to use `Map.of` if you put the condition as a conditional operator to each value. Yet it would make the intention less clear.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/20741#issuecomment-2321574112