Khronos public bugtracker – Bug 763
The description of Table2.4 about Integer Handling is conflict with other contents in this Spec
Last modified: 2017-02-22 09:31:20 PST
In page 24. of OpenGL ES Version 3.0 (August 6, 2012). The description of Table 2.4 said "The “Integer Han-dling” column indicates how fixed-point data types are handled", but I think it should not be fixed-point here, it should be int data types.
In Spec, Page 24 said:
The “Integer Han-dling” column indicates how fixed-point data types are handled:
“integer” means that they remain as integer values,
and “flag” means that they are either converted to floating-point directly, or converted by normalizing to[0;1](for unsigned types) or[ 1;1](for signed types), depending on the setting of the normalizedflag in VertexAttribPointer.
This is conflict with Page 25:
The normalized flag is ignored if type is FIXED, FLOAT, or HALF_FLOAT.
So I think FIXED type should ignore normalized flag, and int type should depend on normalized flag during conversion.
Reassign OpenGL ES bugs in the public Bugzilla to the ES WG Chair so
they'll get more attention.
The wording here actually got tweaked for ES 3.1/3.2 and makes more sense now:
> The “Integer Handling” column in table 10.2 indicates how integer and fixedpoint
> data are handled. “integer” means that they remain as integer values; such
> data are referred to as pure integers. “flag” means that either normalize or cast
> behavior applies, as described below, depending on whether the normalized flag
> to the command is TRUE or FALSE, respectively. normalize means that values are
> converted to floating-point by normalizing to [0, 1] (for unsigned types) or [−1, 1]
> (for signed types), as described in equations 2.1 and 2.2, respectively. cast means
> that values are converted to floating-point directly.
I think the new language fixes this issue appropriately?
By default we're not currently updating the 3.0 spec without an explicit request - and it's been four years since this bug was opened. I'm going to close this bug, but if you want to specifically request a fix in the ES 3.0 spec, please feel free to re-open.