[Mesa-dev] [PATCH 2/5] glsl: Ban #undefining __LINE__ and friends on GLES2.

[Erik Faye-Lund]

On Wed, May 3, 2017 at 1:23 AM, Eric Anholt <eric at anholt.net> wrote:
> Erik Faye-Lund <kusmabite at gmail.com> writes:
>
>> On Tue, May 2, 2017 at 7:36 PM, Eric Anholt <eric at anholt.net> wrote:
>>> Fixes deqp_gles2 undefine_invalid_object_* failures.
>>> ---
>>>  src/compiler/glsl/glcpp/glcpp-parse.y | 7 ++-----
>>>  1 file changed, 2 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/src/compiler/glsl/glcpp/glcpp-parse.y b/src/compiler/glsl/glcpp/glcpp-parse.y
>>> index e113253061f6..5cb2a380605b 100644
>>> --- a/src/compiler/glsl/glcpp/glcpp-parse.y
>>> +++ b/src/compiler/glsl/glcpp/glcpp-parse.y
>>> @@ -284,7 +284,8 @@ control_line_success:
>>>                   *    It is an error to undefine or to redefine a built-in
>>>                   *    (pre-defined) macro name.
>>>                   *
>>> -                 * The GLSL ES 1.00 spec does not contain this text.
>>> +                 * The GLSL ES 1.00 spec does not contain this text, but
>>> +                 * dEQP's preprocess test in GLES2 checks for it.
>>
>> Wouldn't it be better to fix dEQP? Or at least get Khronos to clarify the spec?
>>
>> I mean, the spec is what defines the behavior, not dEQP. There's been
>> a bunch of mistakes in dEQP that has been corrected, why would this be
>> any different?
>
> Khronos regularly redefines old specs to say new things, and sometimes
> they do so without updating the PDFs.  The preprocessor is extremly
> underspecified, so this new GLES3 spec text is likely clarification of
> what old behavior should have been.

Actually, seems there's already at least some form for spec-backing
for this. The wording

"#define and #undef functionality are defined as for C++, for macro
definitions both with and without macro parameters."

from section 3.4 Preprocessor of the OpenGL ES Shading Language 1.00
spec, combined with

"If any of the pre-defined macro names in this subclause, or the
identifier defined, is the subject of a #define or a #undef
preprocessing directive, the behavior is undefined."

from section 16.8.3 [cpp.predefined] of the ISO C++98 (which was the
most recent C++ specification at the time the GLSL ES 1.00 was
written) should also imply that an error is OK.