[Coastrnaut]

Related to #998

With Flag_AlternateNanPropagation, which projectM uses for all preset shaders, the GLSL generator replaces every multiplication with a mult0() helper. The int * int overload returned a float, but in HLSL the product of two ints is an int, and the generated code uses it as one, e.g. mult0(a, b) + 1, mult0(a, b) / 2 or int mask = mult0(a, b). Desktop GLSL converts int to float implicitly, so this mostly went unnoticed there, but GLSL ES has no implicit conversions, and these shaders fail to compile on OpenGL ES:

error C7011: implicit cast from "int" to "float"
error C7011: implicit cast from "float" to "int"

This makes the int overload return an int and adds the missing ivec2/ivec3/ivec4 overloads. Integers have no NaN, so a plain multiplication gives the same result. The float, vector and matrix overloads are unchanged.

How I found it

Master accepts GLES 3.0 contexts since e98fca8, so I compiled all 14,536 presets from cream-of-the-crop, en-d, milkdrop-original and projectm-classic with a GLES build in an OpenGL ES context (NVIDIA, via WGL_EXT_create_context_es2_profile) and compared with desktop GL 3.3. All of projectM's shaders use #version 300 es, so the driver applies GLSL ES 3.00 rules. 10 presets failed only on GLES:

• 3 use sample as an identifier, which GLSL ES reserves. Fix HLSL constructs in Milkdrop presets that failed to transpile #1062 renames it.
• 7 failed with the implicit int → float cast above, e.g. martin - stratospheric turbulences, EVET - Doom Vortex, EVET - Spiracology 2.

martin - funky illusions failed on both desktop and GLES because of the int ← float assignment.

Testing

• GLES: the 8 presets compile now. The only GLES-only failures left are the 3 sample ones that Fix HLSL constructs in Milkdrop presets that failed to transpile #1062 fixes; everything else that fails on GLES fails on desktop as well.
• Desktop: I rendered all 14,536 presets with fixed audio, frame times and random seeds (local patch only, identical on both sides) and the preset folders as texture search path. Compile/link results and load failures are identical to master except that martin - funky illusions now compiles. Rendered one by one, the frames are identical to master except for two presets (martin - gate to moria, martin - mucus cervix) whose generated shaders only differ in the unused helper definitions and which also render differently between runs of master itself.
• New HLSLAlternateMultiplyTest with 3 tests; 2 of them fail on master. All unit tests pass.

The NVIDIA driver always creates a GLES 3.2 context, so this covers the GLSL ES 3.00 language rules, not the 3.0 API limits mentioned in #998 (e.g. rendering to the GL_RG16F motion vector texture).

Retested on current master (2026-11-12)

• Unit tests: 167/167 libprojectM and 98/98 playlist tests pass.
• Presets: I loaded all 14,536 presets from cream-of-the-crop, en-d, milkdrop-original and projectm-classic one after another with this branch and with master (GL 3.3 core, RTX A6000, preset folders as texture search path, 30 frames each with fixed audio, frame times and random seeds; the seeds are a local patch only, identical on both sides). The only compile/link change is martin - funky illusions, which now compiles. All frames before it are identical to master; later ones differ because each preset continues from the state the previous one left behind.
• Together with my other PRs: All my open PRs (Fix calling convention of glGetStringi pointer in GLProbe #1059 to Build for Windows Store (UWP) without app-side workarounds #1065, Return int from the NaN-safe multiplication of two ints #1067 to Fix shaders that declare or name samplers through macros #1072, Build Apple frameworks for iOS, tvOS, watchOS and visionOS, and support EAGL contexts #1075, Don't shift the blur textures by Direct3D's half-pixel offset #1078 and Render with 10 bits per color channel if the target framebuffer has more than 8 #1079) merge with each other and with master without conflicts. Merged together (all except Render with 10 bits per color channel if the target framebuffer has more than 8 #1079, which I opened after this run), they build on desktop GL and GLES, all unit tests pass, and the corpus gives 0 failed loads, 0 shader compile/link failures and 0 GL errors in an OpenGL ES context, with and without textures.