Handled the case where int and float are of correct format, but large. The GLSL spec is not very cl…

Handled the case where int and float are of correct format, but large. The GLSL spec is not very clear on how integers should be interpreted for expressions. C99 says the expression is of type intmax_t. I am parsing all integers as int except those in expressions, which are being parsed as unsigned int.
Committed: https://code.google.com/p/angleproject/source/detail?r=1179

[kbr1, 2012-06-29 17:59:26]

LGTM with a few more tests added.

http://codereview.appspot.com/6351051/diff/1/tests/preprocessor_tests/if_test...
File tests/preprocessor_tests/if_test.cpp (right):

http://codereview.appspot.com/6351051/diff/1/tests/preprocessor_tests/if_test...
tests/preprocessor_tests/if_test.cpp:651: TEST_F(IfTest, IntegerOverflow)
Better named HexIntegerOverflow. Please add tests for octal and decimal ints, as
well as floating-point overflow. Thanks.

Also, please test these constants in the body of shaders as well as preprocessor
directives. That's where we'll see the most usage in practice.

[Alok Priyadarshi, 2012-06-29 19:33:05]

http://codereview.appspot.com/6351051/diff/1/src/compiler/preprocessor/new/Pr...
File src/compiler/preprocessor/new/Preprocessor.cpp (right):

http://codereview.appspot.com/6351051/diff/1/src/compiler/preprocessor/new/Pr...
src/compiler/preprocessor/new/Preprocessor.cpp:98: case Token::CONST_INT:
I do not think overflow check belongs here. It should be done at the compiler
level. The reason I added it here is because the old preprocessor performs this
check. I feel the same way about the check for identifier length. The old
preprocessor complains if it is > 256.

These checks should be performed by the glue code between the preprocessor and
compiler, which is glslang.l. I will move these checks there as soon as we get
rid of the old preprocessor. That is the reason why I did not add tests for 
integer/float overflows in the shader body.

http://codereview.appspot.com/6351051/diff/1/tests/preprocessor_tests/if_test...
File tests/preprocessor_tests/if_test.cpp (right):

http://codereview.appspot.com/6351051/diff/1/tests/preprocessor_tests/if_test...
tests/preprocessor_tests/if_test.cpp:651: TEST_F(IfTest, IntegerOverflow)
On 2012/06/29 17:59:26, kbr1 wrote:
> Better named HexIntegerOverflow. Please add tests for octal and decimal ints,
as

Good idea!

> well as floating-point overflow. Thanks.
> 
> Also, please test these constants in the body of shaders as well as
preprocessor
> directives. That's where we'll see the most usage in practice.

See my other comment about overflow checks for float and constants in the body
of shaders.

[kbr1, 2012-06-30 01:28:25]

http://codereview.appspot.com/6351051/diff/1/src/compiler/preprocessor/new/Pr...
File src/compiler/preprocessor/new/Preprocessor.cpp (right):

http://codereview.appspot.com/6351051/diff/1/src/compiler/preprocessor/new/Pr...
src/compiler/preprocessor/new/Preprocessor.cpp:98: case Token::CONST_INT:
On 2012/06/29 19:33:06, Alok Priyadarshi wrote:
> I do not think overflow check belongs here. It should be done at the compiler
> level. The reason I added it here is because the old preprocessor performs
this
> check. I feel the same way about the check for identifier length. The old
> preprocessor complains if it is > 256.
> 
> These checks should be performed by the glue code between the preprocessor and
> compiler, which is glslang.l. I will move these checks there as soon as we get
> rid of the old preprocessor. That is the reason why I did not add tests for 
> integer/float overflows in the shader body.

I see. I didn't understand this before. Your plan sounds fine. LGTM then modulo
the addition of the other tests I mentioned.