Bug 132464

Summary:

Reduce calls to CFURLCacheCopySharedURLCache

Product:

WebKit

Reporter:

Pratik Solanki <psolanki>

Component:

New Bugs

Assignee:

Pratik Solanki <psolanki>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

ap, beidson, kling, psolanki

Priority:

Keywords:

InRadar

Version:

528+ (Nightly build)

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	none
Patch with assert and raw pointer	ap: review+

Description Pratik Solanki 2014-05-01 23:27:04 PDT

Reduce calls to CFURLCacheCopySharedURLCache

Comment 1 Pratik Solanki 2014-05-01 23:31:28 PDT

Created attachment 230655 [details]
Patch

Comment 2 Pratik Solanki 2014-05-01 23:33:50 PDT

My only concern here would be if someone was calling CFURLCacheSetSharedCache or [NSURLCache setSharedURLCache] after we cache the url ref in the static. I could't find any such callers though.

Comment 3 Alexey Proskuryakov 2014-05-02 00:51:43 PDT

Comment on attachment 230655 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=230655&action=review

> Source/WebKit2/NetworkProcess/mac/NetworkResourceLoaderMac.mm:89
> +    static RetainPtr<CFURLCacheRef> cache = adoptCF(CFURLCacheCopySharedURLCache());

This can't be a RetainPtr, because we don't want exit time destructors. A plain pointer would be appropriate.

An ASSERT in debug builds would ensure that we don't break this optimization, although that may be overkill.

Comment 4 Pratik Solanki 2014-05-02 10:57:49 PDT

Created attachment 230675 [details]
Patch with assert and raw pointer

Comment 5 Alexey Proskuryakov 2014-05-02 11:30:10 PDT

Comment on attachment 230675 [details]
Patch with assert and raw pointer

View in context: https://bugs.webkit.org/attachment.cgi?id=230675&action=review

> Source/WebKit2/NetworkProcess/mac/NetworkResourceLoaderMac.mm:89
> +    static CFURLCacheRef cache = CFURLCacheCopySharedURLCache();

I'd also add ASSERT(isMainThread()) here, as this initialization is not thread safe.

Comment 6 Pratik Solanki 2014-05-03 17:03:35 PDT

<rdar://problem/16806694>

Comment 7 Pratik Solanki 2014-05-03 17:15:25 PDT

Committed r168231: <http://trac.webkit.org/changeset/168231>

Comment 8 Darin Adler 2014-05-03 20:28:27 PDT

Comment on attachment 230675 [details]
Patch with assert and raw pointer

View in context: https://bugs.webkit.org/attachment.cgi?id=230675&action=review

> Source/WebKit2/NetworkProcess/mac/NetworkResourceLoaderMac.mm:93
> +#if !ASSERT_DISABLED
> +    RetainPtr<CFURLCacheRef> currentCache = adoptCF(CFURLCacheCopySharedURLCache());
> +    ASSERT(cache == currentCache.get());
> +#endif

To avoid this ugly if, just get rid of the local variable:

    ASSERT(currentCache == adoptCF(CFURLCacheCopySharedURLCache()));

Also no need for the call to get() when doing ==.

Comment 9 Pratik Solanki 2014-05-04 17:43:44 PDT

(In reply to comment #8)
> (From update of attachment 230675 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=230675&action=review
> 
> > Source/WebKit2/NetworkProcess/mac/NetworkResourceLoaderMac.mm:93
> > +#if !ASSERT_DISABLED
> > +    RetainPtr<CFURLCacheRef> currentCache = adoptCF(CFURLCacheCopySharedURLCache());
> > +    ASSERT(cache == currentCache.get());
> > +#endif
> 
> To avoid this ugly if, just get rid of the local variable:
> 
>     ASSERT(currentCache == adoptCF(CFURLCacheCopySharedURLCache()));
> 
> Also no need for the call to get() when doing ==.

Yes, this looks much better. I landed this in <http://trac.webkit.org/changeset/168250>.