1368571 - Marshal proxies from chrome process to remote clients sans handler

Status: RESOLVED FIXED

(Core :: Disability Access APIs, enhancement)

Description

[(No longer employed by Mozilla) Aaron Klotz]

This is to deal with situations like 32-bit nvda.exe against 64-bit Firefox.

Comment 1

[(No longer employed by Mozilla) Aaron Klotz]

Attachment: Strip handler out of marshaled object references

When I test this against inspect.exe, the handler is loaded in our chrome process but not in inspect.exe itself.

Comment 2

[Jim Mathies [:jimm]]

Comment on attachment 8875083 [details] [diff] [review]
Strip handler out of marshaled object references

Review of attachment 8875083 [details] [diff] [review]:
-----------------------------------------------------------------

Looks great. One question, non-issue based.

::: ipc/mscom/Objref.cpp
@@ +187,5 @@
> +  if (FAILED(hr) || bytesRead != sizeof(numEntries)) {
> +    return false;
> +  }
> +
> +  // We'll try to use a stack buffer if resAddrSize <= kMinDualStringArraySize

I don't understand why you chose two different methods for creating this target buffer. Is the use of the static buf for perf reasons?

Comment 3

[(No longer employed by Mozilla) Aaron Klotz]

(In reply to Jim Mathies [:jimm] from comment #2)
> Comment on attachment 8875083 [details] [diff] [review]
> Strip handler out of marshaled object references
> 
> Review of attachment 8875083 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> Looks great. One question, non-issue based.
> 
> ::: ipc/mscom/Objref.cpp
> @@ +187,5 @@
> > +  if (FAILED(hr) || bytesRead != sizeof(numEntries)) {
> > +    return false;
> > +  }
> > +
> > +  // We'll try to use a stack buffer if resAddrSize <= kMinDualStringArraySize
> 
> I don't understand why you chose two different methods for creating this
> target buffer. Is the use of the static buf for perf reasons?

That is correct. Ideally we will avoid the heap allocation in most cases. From what I've seen, it is most likely that resAddrSize will fall within the kMinDualStringArraySize limit, however that is not guaranteed, hence the heap option.

Comment 4

[(No longer employed by Mozilla) Aaron Klotz]

https://hg.mozilla.org/integration/mozilla-inbound/rev/46699016ed9c06e82c88f4b255246430db2c9dd3
Bug 1368571: Modify handler and interceptor marshaling code so that it strips out any handlers from proxies that are destined for non-Gecko processes; r=jimm

Comment 5

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/46699016ed9c