Flush instruction cache for deserialized code objects.

Flush instruction cache for deserialized code objects.

This fixes the flaky crashes on ARM when running preparser test suite in
optdebug mode.

R=svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=18314

[ulan, 2013-12-16 12:37:23]

PTAL

[Sven Panne, 2013-12-16 12:53:53]

lgtm

https://chromiumcodereview.appspot.com/107543003/diff/1/src/serialize.cc
File src/serialize.cc (right):

https://chromiumcodereview.appspot.com/107543003/diff/1/src/serialize.cc#newc...
src/serialize.cc:799: while (it.has_next()) {
Not directly your code, but I need a place to whine: We should really, really
unify the tons of different iterators and iterator-like interfaces to something
STL-like. We have dozens of useless variations, and the one at hand makes it
basically impossible to use a (more readable) for-loop or something from
<algorithm>. :-(
</EndOfWhining>

"mapM_ flushICache . pages . codeSpace" FTW! ;-)

[ulan, 2013-12-16 12:56:23]

On 2013/12/16 12:53:53, Sven Panne wrote:
> lgtm
> 
> https://chromiumcodereview.appspot.com/107543003/diff/1/src/serialize.cc
> File src/serialize.cc (right):
> 
>
https://chromiumcodereview.appspot.com/107543003/diff/1/src/serialize.cc#newc...
> src/serialize.cc:799: while (it.has_next()) {
> Not directly your code, but I need a place to whine: We should really, really
> unify the tons of different iterators and iterator-like interfaces to
something
> STL-like. We have dozens of useless variations, and the one at hand makes it
> basically impossible to use a (more readable) for-loop or something from
> <algorithm>. :-(
> </EndOfWhining>
> 
> "mapM_ flushICache . pages . codeSpace" FTW! ;-)

Thanks!

> "mapM_ flushICache . pages . codeSpace" FTW! ;-)
should be "mapM_ (flushICache . pages) $ codeSpace" I think :)

[ulan, 2013-12-16 13:08:39]

Committed patchset #1 manually as r18314.