On Mar 3, 2007, at 3:45 PM, Gregory Shimansky wrote:
Geir Magnusson Jr. wrote:
On Feb 15, 2007, at 10:55 AM, Alexey Varlamov wrote:
2007/2/15, Gregory Shimansky <[EMAIL PROTECTED]>:
Alexey Varlamov wrote:
> In my understanding, we face 2 issues here:
> 1) Support "-XX:" format for internal options; this is
straightforward
> and I'll fix this shortly.
> This way such options will not be refused but silently ignored.
Ignoring them silently I think it not the right thing to do. Why
not add
a warning like JRockit does, so that application developers
would pay
more attention to the options that they pass to the VM?
Yes, this should be resolved with the 2nd part; AFAIU this is the
only
way to do it consistently.
Don't you suggest to hardcode all options we aware of at the moment
and warn on uknown? IMO this would be unmaintanable, e.g. jitrino
has
tons of options and they are added/changed quite often.
I suspect the solution will require some refactoring, but you
could iteratively feed each option to each component as they are
initialized, and the component simply reports 1/0 if it cares or
not. Then, for any option for which the core VM didn't care, and
the compomenets didn't care, we report a "unknown option" message.
This may be not so simple because some options may be actually
choosing different components from several possible choices. E.g.
if -Xint is specified, then there is no reason to ask JIT component
about what options it supports.
So before asking components about supported options, VM will have
to determine the set of components it will use and initialize.
Right. What I was trying to say was "as the VM brings up the
components as specified by the configuration, give them the list of
options, and figure out which each cares about, and then at the end,
emit warnings for any option that the VM overall didnt' care about"
geir
--
Gregory
