CSR for posix_spawn on JDK 12

David Holmes david.holmes at oracle.com
Wed Dec 5 11:02:42 UTC 2018


On 5/12/2018 8:54 pm, Alan Bateman wrote:
> On 05/12/2018 10:44, David Holmes wrote:
>>
>>
>> so IMHO for this issue the "specification" should simply be:
>>
>> Update the allowed values of the jdk.lang.Process.launchMechanism 
>> property on Linux to accept the value "posix_spawn", which will use 
>> the posix_spawn() API. The default value of "vfork" remains unchanged.
>>
> That is what CSR JDK-8214511 is about. I agree the wording could have 
> been a lot simpler but there is nothing controversial.

I agree, but there seemed to be some concern over what the 
"specification" should be in this case and how the existing property was 
previously specified.

> To Thomas's question as to why a CSR is needed then it is because it's 
> extending a supported interface. This is different to many of the 
> undocumented options which were added for JDK debugging or testing 
> purposes.

As Joe already stated, because we are changing the set of allowed values 
for a system property, there has to be a CSR.

That said this particular property doesn't seem to actually be 
documented directly. No mention of it in the Process API that I can see. 
It seems a semi-secret means of influencing the implementation. So I can 
understand there can be some confusion here.

David

> -Alan


More information about the core-libs-dev mailing list