Use a struct instead of a union for SandboxInterfaceInfo...

Use a struct instead of a union for SandboxInterfaceInfo

BUG=72496
TEST=see bug



Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=74514

[cpu_(ooo_6.6-7.5), 2011-02-09 22:48:41]

.

[rvargas (doing something else), 2011-02-10 00:23:43]

http://codereview.chromium.org/6483006/diff/1/chrome/common/sandbox_init_wrap...
File chrome/common/sandbox_init_wrapper_win.cc (right):

http://codereview.chromium.org/6483006/diff/1/chrome/common/sandbox_init_wrap...
chrome/common/sandbox_init_wrapper_win.cc:33: return (sandbox::SBOX_ALL_OK ==
target_services_->Init());
Are you counting on other parts of the code to properly set the right interface
for a renderer?. I mean, the previous code was failing if a renderer din't have
taget_services, but this code just returns true in that case.

If we believe this is correct, then we might as well remove all the checks here,
right?

[cpu_(ooo_6.6-7.5), 2011-02-10 18:50:51]

http://codereview.chromium.org/6483006/diff/1/chrome/common/sandbox_init_wrap...
File chrome/common/sandbox_init_wrapper_win.cc (right):

http://codereview.chromium.org/6483006/diff/1/chrome/common/sandbox_init_wrap...
chrome/common/sandbox_init_wrapper_win.cc:33: return (sandbox::SBOX_ALL_OK ==
target_services_->Init());
On 2011/02/10 00:23:43, rvargas wrote:
> Are you counting on other parts of the code to properly set the right
interface
> for a renderer?. I mean, the previous code was failing if a renderer din't
have
> taget_services, but this code just returns true in that case.
> 
> If we believe this is correct, then we might as well remove all the checks
here,
> right?

InitializeSandbox() is called regardless of what process starts. So
target_services_ can be null,  It is the if in line 26-31 that today saves the
bacon, in the sense that the caller does a CHECK() against the result of this
call. 

I think what this function does in mac and linux makes more sense than the
windows one.

I agree that we want to get to the point when this kind of tests of process_type
are not necessary but the code is convoluted so I think we want to take it one
step at a time.

[cpu_(ooo_6.6-7.5), 2011-02-10 20:52:59]

ok, this time with closer behavior to the original.

[rvargas (doing something else), 2011-02-10 20:57:11]

LGTM