Move ParseUnaryExpression into ParserBase and add tests.

Move ParseUnaryExpression into ParserBase and add tests.

This also makes PreParser produce the strict_delete error the same way as
Parser (see test).

R=rossberg@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=20079

[marja, 2014-03-18 18:09:24]

rossberg, ptal

[rossberg, 2014-03-19 11:41:46]

https://codereview.chromium.org/203193004/diff/40001/src/parser.h
File src/parser.h (right):

https://codereview.chromium.org/203193004/diff/40001/src/parser.h#newcode513
src/parser.h:513: bool ShortcutLiteralUnaryExpression(
Nit: not sure I like the name. This is doing constant folding, so maybe rather
call it FoldUnaryExpression (and say "constant-fold" instead of "shortcut" in
the comment.

https://codereview.chromium.org/203193004/diff/40001/src/parser.h#newcode521
src/parser.h:521: bool AddUnaryExpressionTypeFeedback(
Similarly here. I'd perhaps call this RewriteUnaryExpression.

In fact, wouldn't it be possible to merge both these helpers into one function
returning the actual expression directly, such that you can simply do

  return BuildUnaryExpression(expression, op, pos, factory())

in ParseUnaryExpression (after checking the DELETE case)?

[marja, 2014-03-19 12:12:43]

thx for comments; pt another l

https://codereview.chromium.org/203193004/diff/40001/src/parser.h
File src/parser.h (right):

https://codereview.chromium.org/203193004/diff/40001/src/parser.h#newcode521
src/parser.h:521: bool AddUnaryExpressionTypeFeedback(
On 2014/03/19 11:41:46, rossberg wrote:
> Similarly here. I'd perhaps call this RewriteUnaryExpression.
> 
> In fact, wouldn't it be possible to merge both these helpers into one function
> returning the actual expression directly, such that you can simply do
> 
>   return BuildUnaryExpression(expression, op, pos, factory())
> 
> in ParseUnaryExpression (after checking the DELETE case)?

Done.

[rossberg, 2014-03-19 12:28:08]

lgtm

[marja, 2014-03-19 13:42:49]

Committed patchset #6 manually as r20077 (presubmit successful).

[marja, 2014-03-19 14:08:24]

I'm taking the liberty to reuse the old codereview for committing a fixed
version (clang build fixes, basically adding "this->". This is because clang is
stricter about resolving the functions which depend on the template class. You
can see the fix as diff between patch sets 7 and 8.

[marja, 2014-03-19 14:08:55]

Committed patchset #8 manually as r20079 (presubmit successful).