After construction, the Loop is effectively const.
This perturbs the IDs in SPIR-V tests because the body block
is created before generating any of the loop code, rather than
only when the body is first referenced.
It also removes some old code that ancient compilers used to need.
However, the main issue is getting access to hash functions for
unordered_map in portable way.
This simplification is a prelude to eliminating what I appear unnecessary
symbol inserts into tables when tokenizing in the preprecessor, which
show up as taking notable time. (Performance issue.) It also simply makes
the preprocessor easier to understand, which it is badly in need of.
The loop test is always emitted before the loop body.
For do-while loops, use a phi node to track whether we're
on the first loop iteration, and only check the loop test
on the second and subsequent iterations.
For do-while loops, the loop test branch no longer occurs
at the top of the loop, so it must get its own selection
merge instruction.
A block can't be the target of more than one merge instruction.
So when the loop test executes after the body (as in do-while in GLSL)
we need to introduce a dummy block to be the target of the selection
merge just before the loop test conditional branch.
The other arm of the branch exits the loop and hence is the
"break block" exception in the structured control flow rules.
TInputScanner advances its internal indices to the next character at
the end of get(), which means, after reading in the last character
in the user-provided shader string, internal index (currentSource)
will point to the next shader string (currentSource == numSources),
which doesn't exist. Then if a location setting method is called,
we will write to some out-of-bound memory.
A test case for this is "#line 10000\n". The eval() method in CPPline()
will evaluate 10000, but at the same time it reads in the next
token, '\n', and the currentSource will be numSources in TInputScanner.
Then a parseContext.setCurrentLine() is called, we are writing to
out-of-bound memory. Another test case will be "#line 10000 0\n".
Added error output to the preprocessor.
This patch distinguishes preprocessing errors with normal parsing
errors and gives glslangValidator the ability to output preprocessing
errors.
The current line number for the #line directive should be passed
in as parameter to the line directive callback. Without it, we
don't know how many empty lines we should output.
The line argument passed into the lineCallback function is the
literal value of the first argument of the #line directive.
lastLine in DoPreprocessing() should be updated taking into
consideration the different definitions for #line between specs.
Add a test to reveal the bug.
Simplify function calls for extensionsTurnedOn().
Lots of places in the code use extensionsTurnedOn(1, ...). This
patch introduces a new method, extensionTurnedOn(), for testing
if a single extension is turned on.
Lots of places in the code use extensionsTurnedOn(1, ...). This
patch introduces a new method, extensionTurnedOn(), for testing
if a single extension is turned on.
