The correct type for the size of a chunk of memory is size_t.

The correct type for the size of a chunk of memory is size_t.

By using uint32, we have bugs on 64-bit platforms: callers passing in a size_t, will have their size truncated, potentially allocating a smaller
chunk than requested. There are a few places this happens, including on the
receiving ends of IPCs(!)

However, coversely, other callers of the API might directly assign the
memory chunk's length to uint32, leading to a different possible truncation
problem. This is guaraded against by limiting operations internally to
std::numeric_limits<uint32_t> in size for now.

There's some minor cascade effects that make the CL look larger than it is.

BUG=164678
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=175987

[Chris Evans, 2012-12-06 20:57:02]

@darin: for base/*
@jam: for the changes in content/* (looks big but is mechanical)

[jam, 2012-12-06 23:30:29]

we can't send size_t in the ipc messages because that would break when 32 bit
processes talk to 64 bit processes (happens today with nacl on 64 bit, and will
happen in the future with 32 bit plugin processes talking to 64 bit chrome).

[cevans, 2012-12-06 23:32:05]

Ok, pausing to reconsider.

On Thu, Dec 6, 2012 at 3:30 PM, <jam@chromium.org> wrote:

> we can't send size_t in the ipc messages because that would break when 32
> bit
> processes talk to 64 bit processes (happens today with nacl on 64 bit, and
> will
> happen in the future with 32 bit plugin processes talking to 64 bit
> chrome).
>
>
https://codereview.chromium.**org/11446048/<https://codereview.chromium.org/1...
>

[Chris Evans, 2012-12-07 22:28:12]

Splitting out the changes to base/shared_memory*
This CL is now focussed just on those changes.

I'll be sending separate CLs out for the IPC changes. (That work is now headed
in a different direction: I'm fixing IPCs that _already_ incorrectly use size_t)

@darin: for base/*
@jam: for the small tweak in content/

On 2012/12/06 23:32:05, cevans wrote:
> Ok, pausing to reconsider.
> 
> On Thu, Dec 6, 2012 at 3:30 PM, <mailto:jam@chromium.org> wrote:
> 
> > we can't send size_t in the ipc messages because that would break when 32
> > bit
> > processes talk to 64 bit processes (happens today with nacl on 64 bit, and
> > will
> > happen in the future with 32 bit plugin processes talking to 64 bit
> > chrome).
> >
> >
>
https://codereview.chromium.**org/11446048/%3Chttps://codereview.chromium.org...>
> >

[jam, 2012-12-10 22:05:52]

content lgtm

[darin (slow to review), 2013-01-09 23:10:24]

LGTM

I like your comment over IM about switching the limits from uint32 to int.