Bug 102520 – [WK2] REGRESSION: After web process crash a new web process is created for every existing page

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit2

[Carlos Garcia Campos]

Even when process model is SharedSecondaryProcess

[Carlos Garcia Campos]

Created attachment 174691 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 174691 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=174691&action=review

Great catch!

> Source/WebKit2/UIProcess/WebContext.h:223
> +    PassRefPtr<WebProcessProxy> ensureWebProcess();

This function is quite confusing - there is no way one could figure out what it does without reading its code.

I could not come up with a better name or signature for it. Perhaps the logic can just be in WebPageProxy::reattachToWebProcess() inline?

[Carlos Garcia Campos]

(In reply to comment #2)
> (From update of attachment 174691 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=174691&action=review
> 
> Great catch!
> 
> > Source/WebKit2/UIProcess/WebContext.h:223
> > +    PassRefPtr<WebProcessProxy> ensureWebProcess();
> 
> This function is quite confusing - there is no way one could figure out what it does without reading its code.
> 
> I could not come up with a better name or signature for it. Perhaps the logic can just be in WebPageProxy::reattachToWebProcess() inline?

We would still need a way to get the shared process, because ensureSharedWebProcess() doesn't return the process and relaunchProcessIfNecessary is deprecated, right?

[Carlos Garcia Campos]

Created attachment 174816 [details]
Updated patch

Move the logic to WebPageProxy::reattachToWebProcess()

[Alexey Proskuryakov]

Comment on attachment 174816 [details]
Updated patch

View in context: https://bugs.webkit.org/attachment.cgi?id=174816&action=review

> Source/WebKit2/UIProcess/WebContext.h:221
> +    PassRefPtr<WebProcessProxy> getOrCreateSharedWebProcess();

I think that this should still be called ensureSharedWebProcess, and return a plain pointer. There is no passing of ownership, so PassRefPtr is confusing.

[Carlos Garcia Campos]

Committed r135066: <http://trac.webkit.org/changeset/135066>