250666 - Implement the npruntime plugin scriptability API

Status: RESOLVED FIXED

(Core Graveyard :: Plug-ins, defect)

Description

[Johnny Stenback (:jst)]

The URL says it all, I've got an implementation that I'm about to attach.

Comment 1

[Johnny Stenback (:jst)]

Attachment: Implement the npruntime plugin scriptability API

Comment 2

[Brendan Eich [:brendan]]

Comment on attachment 152726 [details] [diff] [review]
Implement the npruntime plugin scriptability API

Stuff I already gave to jst over IRC (sans my own misreadings):

nptypes.h too Win16-based: long for int32 should be exception, not rule?
  int is most commonly 32 bits now
Missing breaks in NPVariantToJSVal String and Object cases
Canonicalize int32 stored in double to avoid 2.0 != 2 in NPVariantToJSVal,
  using JS_NewNumberValue in the NPVariantType_Double case
else after return in GetProperty
return ok == JS_TRUE; => return ok; in nsJSObjWrapper::NP_Invoke
return ok && found == JSVAL_TRUE;
  => return ok && !JSVAL_IS_VOID(found); in nsJSObjWrapper::NP_HasProperty
assert that id is int if not string in all the hooks that take an id arg

r+sr=me with these fixes.  I reviewed quickly (food poisoning), but it looks
good to go.

/be

Comment 3

[Brendan Eich [:brendan]]

Comment on attachment 152726 [details] [diff] [review]
Implement the npruntime plugin scriptability API

Nit: JSObjNPPHashEntry seems unused.

Idea: A GC root per nsJSObjWrapper/JSObjWrapperHashEntry makes for a separately
new/delete'd objct (nsJSObjWrapper), the sJSObjWrappers double-hashtable entry,
and the GC roots double-hashtable entry.  You could get rid of the last by
marking each obj reached from the next-to-last, using a GCCallback for
JSGC_MARK_END.

/be

Comment 4

[Asa Dotzler [:asa]]

Comment on attachment 152726 [details] [diff] [review]
Implement the npruntime plugin scriptability API

a=asa (on behalf of drivers) for checkin to 1.8a2.

Comment 5

[David Hallowell]

Nominating for firefox 1.0, hopefully this will be a popular release and it
would certainly help adoption for this new API if it's supported in 1.0.

Not sure how much testing it will need though.

Comment 6

[Johnny Stenback (:jst)]

Ok, after a long day of build bustage across the board thanks to C99, this is
checked in on the trunk. Brendan, I chose to not start changing how we control
the lifetime of JSObjects here for the initial landing, I filed bug 251149 on
tracking that. Marking this FIXED.

Comment 7

[Mike Kaply [:mkaply]]

The change at line 1520 in ns4xPluginInstance.cpp was not correct:

PR_BEGIN_EXTERN_C
      void _releaseobject(NPObject *obj);
PR_END_EXTERN_C

In ns4xPlugin.h, _releaseobject as defined as NP_EXPORT and you can't add
NP_EXPORT to the statement in ns4xPluginInstance.h, since it is defined in
ns4xPlugin.h.

ns4xPluginInstance.cpp should be including ns4xPlugin.h, not definining a custom
reference for _releaseobject.

Comment 8

[Johnny Stenback (:jst)]

Duh, yeah, leftover cruft from hacking. Fixed.

Comment 9

[Ryan Gammon]

Anyone know if there's a plugin sdk with updated versions of np_entry.cpp,
npn_gate.cpp, npp_gate.cpp for the new npruntime stuff (or plans in the works to
do this)?