This is based on spv.paramMemory.frag.out which exercises the
writeonly storage image case.
This appears to need desktop GLSL.
The generated SPIR-V fails validation because the image_write function
takes a parameter which is pointer to an OpTypeImage with Unknown format.
But the parameters passed in are pointer to OpTypeImage with formats Rgba32f
and Rgba16f. The validator rejects this, saying the parameter types must
match.
There's a statement that intends to generate a 32-bit hashcode, but due
to integer promotion, the intermediate values can trigger signed integer
overflow, which is undefined behavior.
To avoid this, cast at least one operand to unsigned int before
multiplying, which will cause the result to be promoted to unsigned int
instead of signed int.
With this patch, I'm able to build core for qemu-x64 with host_asan-ubsan.
Fixed: 60128
Change-Id: Idd644e534116bf29dca8013936ac39901bbe68fc
Reviewed-on: https://fuchsia-review.googlesource.com/c/third_party/glslang/+/428254
Reviewed-by: John Bauman <jbauman@google.com>
Co-authored-by: Drew Fisher <zarvox@google.com>
The glslang public includes dir contains headers with implicit
conversion issues. Add -Wno-conversion to glslang's public config.
Bug: 60140
Bug: 58162
Change-Id: Iec27cb4242e9fdceddd6a3e02044a0bccfa0ce36
Reviewed-on: https://fuchsia-review.googlesource.com/c/third_party/glslang/+/429054
Reviewed-by: Petr Hosek <phosek@google.com>
Co-authored-by: Shai Barack <shayba@google.com>
Add more builder functions to OpExecutionMode, OpExecutionModeId,
OpDecorateString, OpMemberDecorateString.
According to SPIR-V, OpExecutionMode and OpExecutionModeId could
take variable extra operands. Current implementation doesn't support
this and assumes at most 3 operands are extra operands. It is not
true. Similarly, OpDecorateString and OpMemberDecorateString could
support multiple strings either as literal strings or as string
operands. Further, OpDecorate and OpDecorateId have the same problem,
taking variable extra operands.
* Preprocessor related fix
1). Accoding to ESSL spec : All macro names containing two consecutive underscores ( __ ) are reserved for future use as predefined macro names, so just report a warning instead of error when the shader defines the macro names begining with '__';
2. According to spec: If an implementation does not recognize the tokens following #pragma, then it will ignore that pragma, so report a compile-time warning intead of error for the following statement:
#pragma debug(1.23)
3. The 'defined' macro should be allowed to expand and '__LINE__' should be allowed to be replaced with its original line number (otherwise, other expanding macros may change this value).
4. Add a flag 'indentifierSeen' in PPContext to indicate whether the any non-preprocessor tokens is existed before the extension directives, because the built-in symbols and functions are parsed before paring the user shader, so add a 'shaderSource' flag to check this error only for the user shader source;
5. Add missing type int16 and uint16.
* Add test results, remove restriction of #extension.
1. Remove extension restriction in first line , as this is contraversy now.
2. The following shader is compiled failed as glslang consider the keyword 'defined' can not be undefined(in the 9th line: "#define defined BBB")
The shader is as following:
According to ES3.0 spec: It is an error to undefine or to redefine a built-in (pre-defined) macro name.
This rule is aimed to the __LINE__, __FILE__, __VERSION__ and GL_ES,
the keyword "defined" should not be restricted by this rule,
so change the compile error to warning and make the following shader compile successfully.
* 1. Using relaxedError to control error/warning report level. 2. remove #extension restriction. 3. Fix version related issue.
1. Using relaxedError to control error/warning report level. 2. remove #extension restriction. 3. Fix version related issue.
* Add test results
* Turn conditional warnings about pragma to unconditional ones.
According to the extension SPV_GOOGLE_decorate_string,
OpDecorateString (or OpMemberDecorateString) ought to be capable of
supporting multiple literal strings. Each literal strings are padded
with null terminator to make word alignment. The layout is:
Inst | Target | Decoration | Literal String, Literal String, ...
* Remove image2DShadow and other 3 tokens. Refine codes.
Remove image2DShadow and other 3 tokens. Refine codes.
* 110scope.vert has redefinition part of what's removed from 100scope.vert
The current implementation makes tests fail when it is imported from a
different Bazel workspace. We fix that by using
the --test-root flag to pass the rootpath to the
tests.
We need separate concepts for
- total set of extensions ever enabled, for the back end
- current state of extensions, for parsing
- the set of features currently enabled for building the AST
This detangles incorrect conflation of HLSL with GLSL extensions,
defers asking expensive questions until it's time to ask, and removes
some dead code.
The scalar type was already the basic type passed in.
Also factored out of this the checking of extensions for 8/16-bit stuff.
This code seems wrong in several ways, but for now just documenting it.
