REGRESSION (r248440): MiniBrowser no longer loads file:// URLs on launch

[Per Arne Vollan]

This is a regression from https://trac.webkit.org/changeset/248440/webkit.

[Per Arne Vollan]

rdar://problem/54168065

[Brent Fulgham]

Is this a race condition where we don't have the PID yet when we try to create the sandbox extension?

[Per Arne Vollan]

(In reply to Brent Fulgham from comment #2)
> Is this a race condition where we don't have the PID yet when we try to
> create the sandbox extension?

I think the pid is valid, but I will double-check.

[Per Arne Vollan]

Created attachment 376109 [details]
Patch

[Brent Fulgham]

Comment on attachment 376109 [details]
Patch

Oh! Got it.

[Per Arne Vollan]

(In reply to Brent Fulgham from comment #2)
> Is this a race condition where we don't have the PID yet when we try to
> create the sandbox extension?

You are right, Brent. There seems to be a race between creating the sandbox extension when starting the load and getting the pid of the new WebContent process.

[Per Arne Vollan]

Created attachment 376162 [details]
Patch

[Brent Fulgham]

Comment on attachment 376162 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=376162&action=review

> Source/WebKit/Shared/Cocoa/SandboxExtensionCocoa.mm:352
> +        return false;

Do we cleanly recover if this returns false? I.e., if the pid isn't available and so we don't create the extension, does WebKit try again later?

[Per Arne Vollan]

(In reply to Brent Fulgham from comment #8)
> Comment on attachment 376162 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=376162&action=review
> 
> > Source/WebKit/Shared/Cocoa/SandboxExtensionCocoa.mm:352
> > +        return false;
> 
> Do we cleanly recover if this returns false? I.e., if the pid isn't
> available and so we don't create the extension, does WebKit try again later?

If the pid is invalid, the patch falls back to creating an extension without the pid. I am not sure if this is the optimal solution. I am also looking into creating the extension with an audit token instead of a pid.

Thanks for reviewing!

[Per Arne Vollan]

*** This bug has been marked as a duplicate of bug 200772 ***