| Age | Commit message (Collapse) | Author |
|---|
|
|
|
The IR reader does not expect commas.
|
|
Since GLSL permits arrays of structures, we need to store each element
as an ir_constant*, not just ir_constant_data.
Fixes parser tests const-array-01.frag, const-array-03.frag,
const-array-04.frag, const-array-05.frag, though 03 and 04 generate the
wrong code.
|
|
|
|
Implicit conversions were not being performed, nor was there any
type checking - it was possible to have, say, var->type == float
and var->constant_value->type == int. Later use of the constant
expression would trigger an assertion.
Fixes piglit test const-implicit-conversion.frag.
|
|
Also make ir_unop_i2f only operate on signed integers.
|
|
The ir_constant::has_value method already does this.
|
|
|
|
|
|
This is an invasive set of changes. Each user shader tracks a set of other
shaders that contain built-in functions. During compilation, function
prototypes are imported from these shaders. During linking, the
shaders are linked with these built-in-function shaders just like with
any other shader.
|
|
|
|
To quiet a compiler warning.
|
|
In both the preprocessor and in the compiler proper, we use a custom
yyltype struct to allow tracking the source-string number in addition
to line and column. However, we were previously relying on bison's
default initialization of the yyltype struct which of course is not
aware of the source field and leaves it uninitialized.
We fix this by defining our own YYLLOC_DEFAULT macro expanding on the
default version (as appears in the bison manual) and adding
initialization of yylloc.source.
|
|
Previously, any occurence of the unary plus operator would trigger a
bogus type mismatch error. Fix this by making the ast_plus case look
more like the ast_neg case as far as type-checking is concerned.
With this change the shaders/CorrectPreprocess8.frag test in piglit
now passes.
|
|
|
|
The lowering code should probably be moved elsewhere.
|
|
|
|
Coming changes to the handling of built-in functions necessitate this.
|
|
This is quite a large patch because breaking it into smaller pieces
would result in the tree being intermitently broken. The big changes
are:
* Add the ir_var_temporary variable mode
* Change the ir_variable constructor to take the mode as a
parameter and correctly specify the mode for all ir_varables.
* Change the linker to not cross validate ir_var_temporary
variables.
* Change the linker to pull all ir_var_temporary variables from
global scope into 'main'.
|
|
This caught the failure in cloning of ir_dereference_record.
|
|
We're good at propagating error types around, but finding when the
first one was triggered can be painful if we aren't paying attention.
|
|
Otherwise, after linking and freeing the old data, the pointer would
dangle. Partial fix for glsl1-struct*.
|
|
Since the types are singletons across the lifetime of the compiler,
repeatedly compiling a program with the same structure type defined
would drop a copy of the array on the floor per compile.
This is a bit tricky because the static GLSL types are not called with
the talloc-based new, so we have to use the global type context, which
may not be initialized yet.
|
|
We regularly do lookups on the field names of the structure to find
the types within the struct, so returning a structure type with bad
names will lead to lots of error types being found.
|
|
Fixes piglit test glsl-array-length, and provides proper error messages
for negative piglit tests array-length-110.frag, array-length-unsized.frag,
and array-length-args.frag.
|
|
This assertion is triggered by method calls (i.e. array.length()), where
subexpressions[1] is an ast_function_call expression. Since the
assertion itself had a comment saying it could be removed eventually,
simply do so.
Causes negative glslparser tests array-length-110.frag,
array-length-args.frag, and array-length-unsized.frag to pass, but only
because the length() method is not supported yet.
|
|
Fixes piglit test unsized-array-non-const-index.vert.
|
|
ir_dereference_variable always references an ir_variable, so there's no
point in calling a function and NULL-checking the result.
|
|
In ir_expression's signature, I replaced ir->operands[i] with op[i] as
it is more concise; an assertion already ensures these are equal.
|
|
The constant_expression_wrapper was already the only external API, and
much of the internal code used it anyway. Also, it wouldn't ever visit
non-rvalue ir_instructions, so using a visitor seemed a bit unnecessary.
This uses "ir_foo *ir = this;" lines to avoid code churn. These should
be removed.
|
|
This prevents top-level callers from asking for the value of something
that is guaranteed not to have one.
|
|
Consider this test case:
#define EMPTY
int foo = 1+EMPTY+4;
The expression should compile as the sequence of tokens 1, PLUS,
UNARY_POSITIVE, 4. But glcpp has been failing for this case since it
results in the string "1++4" which a compiler correctly sees as a
syntax error, (1, POST_INCREMENT, 4).
We fix this by changing any macro with an empty definition to result
in a single SPACE token rather than nothing. This then gives "1+ +4"
which compiles correctly.
This commit does touch up the two existing test cases which already
have empty macros, (to add the space to the expected result).
It also adds a new test case to exercise the above scenario.
|
|
Without this, the compiler was legitimately complaining about missing
declarations for all of the functions being defined here.
|
|
This quiets warnings about missing declarations otherwise.
|
|
We define the YY_NO_INPUT macro to avoid one needless function being
generated.
for the other needless functions, (yyunput and yy_top_state), we add a
new UNREACHABLE start condition and call these functions from an
action there. This doesn't change functionality at all, (since we
never enter the UNREACHABLE start condition), but makes the compiler
stop complaining about these two functions being defined but not used.
|
|
It's really a bug in flex that these functions are generated with neither
a declaration nor the 'static' keyword, but we can at least avoid the
warnings this way.
|
|
Previously, if the outer #ifdef/#ifndef evaluated to false, the inner
directive would not be parsed correctly, (the identifier as the subject
of the #ifdef/#ifndef would inadvertently be skipped along with the other
content correctly being skipped).
We fix this by setting the lexing_if state in each case here.
We also add a new test to the test suite to ensure that this case is tested.
|
|
And add a test case to ensure that this works.
|
|
By taking advantage of the recently-added hash_table_remove function.
With this change, all existing tests are now valgrind-clean.
|
|
To allow for the removal of a single element from a hash table.
|
|
As it turns out, 4 of our current tests are not valgrind clean,
(use after free errors or so), so this will be helpful for
investigating and fixing those.
|
|
And report PASS or FAIL for each test along the way as well.
|
|
This is convenient for testing the preprocessor independent of the rest of
mesa, (just run glcpp-test in the src/glsl/glcpp/tests).
|
|
These were only ever intended to exist in the original, standalone
implementation of glcpp, (with the idea of dropping them as soon as
the code moved into mesa). The current build system wasn't compiling
this C file, but the presence of the header files could cause problems
if the two copies diverge in the future.
We head those problems off by deleting al of these redundant files.
|
|
The parameters[i] is our inlined variables representing the
parameters, so they are always ir_var_auto. Walk the signature params
in handling "out" values like we do for "in" values to find the mode.
Fixes (with the previous 2 commits):
glsl1-function call with in, out params
glsl1-function call with inout params
|
|
|
|
If they have a return value, this means putting it into a temporary
and making a deref of the temp be the rvalue, since we don't know if
the rvalue will be used or not.
|
|
Fixes:
glsl1-built-in constants
|
|
I'd flipped around the order of two operations in paren-balancing
adventures, and left out the multiply by sign(x) required for negative x.
Fixes:
glsl1-acos(vec4) function
glsl1-asin(vec4) function
glsl1-atan(vec4) function
|
|
Fixes:
glsl1-vector relational (bvec2 ==,!=)
glsl1-vector relational (vec4 !=)
|
