On Wed, Jan 13, 2010 at 9:28 PM, Peter Kasting <[email protected]> wrote:
> On Wed, Jan 13, 2010 at 11:22 AM, Craig Schlenter
> <[email protected]> wrote:
>>
>> Other than the immediate gain of "hiding" crbug.com/28749,
>
> Which you have a patch (actually two, but one with r+) for, right?

True. I was tweaking the latest patch though and was still hoping to
get some feedback on it (thanks for the alignment comments btw.).

>> I think the biggest
>> benefit is that end users relying on 4.4 builds are likely to have a more
>> stable
>> experience in future since it is a safe default.
>
> I don't understand what you mean.  Whichever way we set the flag, we'll be
> compiling with it.  So that will be the "safe" thing for others to compile
> us with.  Throwing -fstrict-aliasing has the added advantage that it makes
> _both_ settings "safe".

It does however make the default compile "unsafe" for the case where
the compiler
doesn't throw a warning as is the case for issue 28749.

>> The other thing it buys us is a more relaxed timetable to solve the
>> aliasing
>> problems if it doesn't break the tree by default.
>
> Do we have other aliasing problems?  If so my opinion changes.

There are some aliasing issues still in play. Off the top of my head:

1. unit_tests has what might be an issue in stl_tree.h or a compiler issue
 (bug filed with gcc ppl) and it was already using
-fno-strict-aliasing to avoid this with gcc 4.4.
2. skia has an unresolved issue. Note the filing date:
http://code.google.com/p/skia/issues/detail?id=18
3. v8 has an aliasing issue in dtoa and somewhere else that I forget.
4. there are issues in third party code of course eg. icu, webkit etc.

--Craig
-- 
Chromium Developers mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-dev

Reply via email to