Bug 173145

Summary: Add API::IconDatabaseClient
Product: WebKit Reporter: Carlos Garcia Campos <cgarcia>
Component: WebKit2Assignee: Nobody <webkit-unassigned>
Status: RESOLVED FIXED    
Severity: Normal CC: achristensen, andersca, darin
Priority: P2    
Version: WebKit Nightly Build   
Hardware: Unspecified   
OS: Unspecified   
Bug Depends on:    
Bug Blocks: 173146    
Attachments:
Description Flags
Patch achristensen: review+

Description Carlos Garcia Campos 2017-06-09 03:30:34 PDT
We will use this instead of the C API in GTK+.
Comment 1 Carlos Garcia Campos 2017-06-09 03:32:54 PDT
Created attachment 312418 [details]
Patch
Comment 2 Alex Christensen 2017-06-09 10:24:40 PDT
Comment on attachment 312418 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312418&action=review

> Source/WebKit2/UIProcess/WebIconDatabase.h:118
> +    std::unique_ptr<API::IconDatabaseClient> m_client;

I think this is wrong.  Usually we have the API object own the WebKit namespace object.
Comment 3 Carlos Garcia Campos 2017-06-10 01:37:56 PDT
Comment on attachment 312418 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312418&action=review

>> Source/WebKit2/UIProcess/WebIconDatabase.h:118
>> +    std::unique_ptr<API::IconDatabaseClient> m_client;
> 
> I think this is wrong.  Usually we have the API object own the WebKit namespace object.

Could you point to an example? I think I've followed what all other objects do. 

 - WebPageProxy in WebKit namespace owns API::LoaderClient, API::PolicyClient, API::NavigationClient, and a lot more clients
 - WebProcessPool in WebKit namespace owns API::AutomationClient, API::DownloadClient, API::LegacyContextHistoryClient and API::CustomProtocolManagerClient

or did I misunderstand what you mean?
Comment 4 Alex Christensen 2017-06-12 09:59:28 PDT
Comment on attachment 312418 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312418&action=review

> Source/WebKit2/UIProcess/API/APIIconDatabaseClient.h:2
> + * Copyright (C) 2014 Apple Inc. All rights reserved.

2017

>>> Source/WebKit2/UIProcess/WebIconDatabase.h:118
>>> +    std::unique_ptr<API::IconDatabaseClient> m_client;
>> 
>> I think this is wrong.  Usually we have the API object own the WebKit namespace object.
> 
> Could you point to an example? I think I've followed what all other objects do. 
> 
>  - WebPageProxy in WebKit namespace owns API::LoaderClient, API::PolicyClient, API::NavigationClient, and a lot more clients
>  - WebProcessPool in WebKit namespace owns API::AutomationClient, API::DownloadClient, API::LegacyContextHistoryClient and API::CustomProtocolManagerClient
> 
> or did I misunderstand what you mean?

After talking to some people, I now think this is correct.  Sorry about the misunderstanding.  API namespace objects own WebKit namespace objects, but clients pass calls in the opposite direction.  This is correct.
We should us a UniqueRef here because it's never null.
Comment 5 Carlos Garcia Campos 2017-06-12 10:04:47 PDT
(In reply to Alex Christensen from comment #4)
> Comment on attachment 312418 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=312418&action=review
> 
> > Source/WebKit2/UIProcess/API/APIIconDatabaseClient.h:2
> > + * Copyright (C) 2014 Apple Inc. All rights reserved.
> 
> 2017
> 
> >>> Source/WebKit2/UIProcess/WebIconDatabase.h:118
> >>> +    std::unique_ptr<API::IconDatabaseClient> m_client;
> >> 
> >> I think this is wrong.  Usually we have the API object own the WebKit namespace object.
> > 
> > Could you point to an example? I think I've followed what all other objects do. 
> > 
> >  - WebPageProxy in WebKit namespace owns API::LoaderClient, API::PolicyClient, API::NavigationClient, and a lot more clients
> >  - WebProcessPool in WebKit namespace owns API::AutomationClient, API::DownloadClient, API::LegacyContextHistoryClient and API::CustomProtocolManagerClient
> > 
> > or did I misunderstand what you mean?
> 
> After talking to some people, I now think this is correct.  Sorry about the
> misunderstanding.

np, the whole thing is indeed confusing at first.

>  API namespace objects own WebKit namespace objects, but
> clients pass calls in the opposite direction.  This is correct.
> We should us a UniqueRef here because it's never null.

Didn't know UniqueRef, do you want me to change this patch before landing to use UniqueRef, or better do that in a follow up and switch all clients to UniqueRef?
Comment 6 Alex Christensen 2017-06-12 10:11:50 PDT
Switching all clients to UniqueRef would be nice.
Comment 7 Carlos Garcia Campos 2017-06-12 10:13:40 PDT
(In reply to Alex Christensen from comment #6)
> Switching all clients to UniqueRef would be nice.

Then I'll keep this simple and switch all clients in a follow up.
Comment 8 Carlos Garcia Campos 2017-06-12 10:19:35 PDT
Committed r218107: <http://trac.webkit.org/changeset/218107>