Emulator

you should see if this worked in a build before 4.0-1288.

Also make sure you don't have disable destination alpha disabled.

s/disabled/checked. Just, y'know, keep that checkbox empty.

I tested with older builds, first with 4.0-1146, which is the last one before TEV Fixes New was merged. No change. Then I felt the temptation and tried r6505 (the oldest build I had archived), but at least OGL was so glitched (and crashed on save game loading) on my machine that I had to pass. Then I randomly chose a build from 3.5 era, namely 3.5-1066, and the glitch seems exactly the same. So I make now the assumption that this has never worked on Intel HD.

What came to my mind when I looked at the introduction scene and then the scene where I get the star, is that the problem seems to be in the reflections. Water gets the color and the reflection clearly wrong, as proved by these images. But I'm not saying this is Dolphins fault, since it only happens with Intel HD. Besides, Intel has made the OpenGL support better with each driver update, this might get fixed too, in the future.

http://imgur.com/j7FAIkR

http://imgur.com/9X1mKy7

And yes, I did these tests in portable mode, default OpenGL settings.

Delroth's right, this is not related to PR #509. I have tested this with a build that has the fix for that issue, and it does not fix it. Actually, the #509 fix only affects D3D (that is glitched). I tried to build a version that has the problematic bit shift changed with BOTH backends, and OpenGL fails to start a game. What I understood from the error description it has something to do with OpenGL not being able to do the comparison >= 128 ? 1 : 0 with a interger that has form int(x,y,z). But I really don't know if this is the problem.
Anyway, 4.0-1146 that I tested does not have the commit that caused the whole PR #509 issue, so...

But I will try 3.5-663, thank you for the tip!

Woah! I tested 3.5-663, and even if it was the glitchiest experience I've seen in a while, the water actually rendered like it should! There were a lot of colored polygons flickering around the screen, this is something that apparently was fixed between 3.5-663 and 3.5-1066 when OpenGL got some big updates. But some of these updates also broke the water effect for Intel HD, and I guess I just have to fork the commit :) Thanks again for the tip!

Here are some pictures:

http://imgur.com/jTKRXzs (perfect)
http://imgur.com/i8hZ882 (messed up graphics, but still playable)
http://imgur.com/RjSWDDe (pretty distracting occasionally)
http://imgur.com/7kZ8mX7 (cursor movement makes it worse here)

The better option would be trying to get Intel to fix their forkin' drivers.

...just as I was afraid of, next I tried 3.5-1025, which is the next release developement build after 3.5-663. Guess what, the water effect gets broken and also the major ugly polygon-glitches get fixed. No less than 1025-663=362 commits in between. And no builds.

But there you have it, this 3.5-1025 (Merge branch 'GLSL-master', March 15 2013, On Github found on page 107 of commits (today), commit code e1a081ad2dee7127e6c1fe2827eb1eb59ca24c6e) is The Merge that brought this little annoyance for us Intel HD users. Maybe I'll do some building next, to find The Commit :)

Except... It's merged as one single commit, and I can't find the GLSL-master -branch, is it still somewhere?

JMC: You're so right. I'll stop forking here.

if you are curious, because it would help in narrowing down what intel is doing wrong, you can probably find the GLSL branch on here: https://dl.dolphin-emu.org/builds/

Ok I couldn't resist the challenge. Unfortunately, and because of the nature of GLSL branch, the culprit commit for this glitch seems to be somewhat untraceable.

I went through the GLSL-master x64-builds found at the url JMC4789 posted above. It seems that OpenGL backend is "broken" for my Intel HD 4600 up to build dolphin-GLSL-master-3.5-617-x64 . Most of the builds I tested (starting from dolphin-GLSL-master-3.0-947-x64) show only black screen, other show unicolored back-/foreground, no real visual output. Then the first build I got to show something is dolphin-GLSL-master-3.5-619-x64. Mario's shadow is "blocky" and lighting seems to dark, but SMG1 runs and is playable. Unfortunately the water and its reflections are messed already in this build. There's like 700 commits in between branch creation and this.

So the next step is to revert each commit one by one from this first working build, right? ;)

I'd say this is as narrowed as we can get. Won't help Intel much, personally I'm okay just checking the situation after each new Intel HD driver release.

Intel released new drivers a couple of days ago (version 15.36.3.64.3907), and this is just an update. Unfortunately the new drivers break OpenGL on Intel HD altogether, so this issue can't be tested (today) with recent builds, but builds up until 4.0-575 work, and they show the same issue with water rendering / reflecting wrong.

So the current drivers do not fix this.

hosode: We have worked around this new intel black screen issue, so you should be able to test this on current master again.

Tested with the latest build (4.0-3367), portable mode, default settings. I only copied the savegame (not savestate) from my main installation. OGL works now, which is great of course :)
But the issue is still there. The more I look at it the more convinced I am that the main problem is in the way water reflects things.
I agree this being marked as questionable. If Intel HD really seems to be the only platform struggling with this (non-game-breaking) issue and therefore the problem being most likely in the drivers, then wontfix is also very acceptable.

Waiting on a verdict from a graphics guy.

Has there been any new driver releases?

Yes, there was a driver update which at least has fixed the buffer_storage issue.

Yes I tested SMG1 with the latest drivers that were released less than two weeks ago. As said, it specifically fixed the OGL buffer storage problem (from ~4.0-500), but did not affect this water reflection issue, which tracks down further to OGL rewrite at 3.5-1025.
I have not reported this issue to Intel as this seems so minor that I just did not feel the need. Although they have been very positively responsive with the two (more severe) issues I reported.

It's definitely an Intel HD issue. You may want to report it to them if they care; it'd be nice to have it fixed.

Ok will do!

Hey! As of today, I could not bring myself reporting this issue to Intel, sorry about that.
But I just noticed that there were new Intel HD drivers available on their site (driver version 15.36.14.64.4080).
So first I downloaded the very latest Dolphin development build (4.0-5238) and tested (portable version, default settings) if the problem had resolved with the awesome recent updates. Turns out that some of them broke OpenGL backend in SMG1 for Intel HD altogether, as I got black screen hang when entering Loop-de-loop galaxy. I did not bisect the culprit, since I still hadn't installed the latest drivers.
Well, next I updated the drivers and

1. SMG1 Loop-de-loop galaxy launches on OpenGL and
2. THE ORIGINAL ISSUE IS RESOLVED, water now renders bright blue with fishes and all, just like it should.
   I consider this fixed.