Bug 162929 – [Bindings] Declare dictionary / enumeration template specializations in the header

• Component: Bindings
• Status: RESOLVED FIXED

[Chris Dumez]

Move convertDictionary<>() template specializations to the header in order to:
- Allow them to be used from custom bindings code.
- Prepare for moving dictionaries to their own IDL files (Bug 162912).

[Chris Dumez]

Created attachment 290643 [details]
Patch

[Chris Dumez]

Created attachment 290645 [details]
Patch

[Sam Weinig]

Do the implementations really have to go in the header? Can't we just forward declare what we want to call?

[Chris Dumez]

(In reply to comment #3)
> Do the implementations really have to go in the header? Can't we just
> forward declare what we want to call?

Your comment only applies to jsStringWithCache() right? All other functions are templated so I believe they will need to be defined in the header if we want them to be callable from other compilation units.

I do think we can move the definition of jsStringWithCache() to the cpp though and only keep its declaration in the header.

[Chris Dumez]

(In reply to comment #4)
> (In reply to comment #3)
> > Do the implementations really have to go in the header? Can't we just
> > forward declare what we want to call?
> 
> Your comment only applies to jsStringWithCache() right? All other functions
> are templated so I believe they will need to be defined in the header if we
> want them to be callable from other compilation units.
> 
> I do think we can move the definition of jsStringWithCache() to the cpp
> though and only keep its declaration in the header.

Although I don't know about the performance impact of doing so given that the function was likely being inlined.

[Chris Dumez]

Created attachment 290665 [details]
Patch

[Chris Dumez]

I have moved jsStringWithCache() implementation back to the cpp. I haven't tried instantiating the template functions in the header and moving the template specializations to the cpp yet.

[Chris Dumez]

(In reply to comment #7)
> I haven't
> tried instantiating the template functions in the header and moving the
> template specializations to the cpp yet.

The thing is that I worry it will lead to linking errors when actually calling the specialization from another compilation unit (which we do not do yet so I won't notice) or about getting weird compilation trouble on Windows.

When I have tried tricks like this in the past, it usually hasn't gone well :/

Do you mind if we try doing so in a follow-up, once we actually start sharing dictionaries (i.e. after Bug 161932 is fixed)?

[Chris Dumez]

(In reply to comment #8)
> (In reply to comment #7)
> > I haven't
> > tried instantiating the template functions in the header and moving the
> > template specializations to the cpp yet.
> 
> The thing is that I worry it will lead to linking errors when actually
> calling the specialization from another compilation unit (which we do not do
> yet so I won't notice) or about getting weird compilation trouble on Windows.
> 
> When I have tried tricks like this in the past, it usually hasn't gone well
> :/
> 
> Do you mind if we try doing so in a follow-up, once we actually start
> sharing dictionaries (i.e. after Bug 161932 is fixed)?

Oh, it looks like it is easier than I thought. I'll try and do it in this patch.

[Chris Dumez]

Created attachment 290688 [details]
Patch

[Chris Dumez]

Comment on attachment 290688 [details]
Patch

Clearing flags on attachment: 290688

Committed r206812: <http://trac.webkit.org/changeset/206812>

[Chris Dumez]

All reviewed patches have been landed.  Closing bug.