Melty Blood Actress Again (SLPM-55184) - Videos no longer function

[GoogleCodeExporter]

1) Please list the latest versions (Official Releases) and/or svn revisions
the game is known work correctly with.

The game works perfectly normally, and always has. This is also a rare game
whose videos actually worked perfectly normally, but here's a twist: I
first noticed the videos stopped working as of r3121.

A bit of quick regression testing against the official public binary
(r3113) keeps the freeze.

I'd hoped it'd be fixed with r3066 which I had laying around; sadly, it did
not. The next most recent version I had compiled was r2655 - where it does
work. Unfortunately, this 400-revision gap is likely too big to investigate
as-is, so I will leave some further details and perhaps a dev can tell me
what file/plugin/etc. to whittle/narrow it down to, so I can begin testing
revisions.

Again, the Videos *DID WORK* before. They're broke now. This is unusual,
considering more often than not, the videos are already broken due to the
usual IPU "fun." Ergo, I am filing this under regression. 

2) What version of the product are you using? On what operating system?

Problem noticed on r3121; whatever broke it occured somewhere between r2655
and r3066.

3) Describe the regression: What is the expected output?  What do you
experience instead?

Expected output, of course, is the videos play. The reality is it locks up.
Console log prints out a lot of this:

sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=005dcec0:D1_MADR=005dced0:D1_QWC=00000000>
    <D2_CHCR=00000101:D2_TADR=006ed130:D2_MADR=004be390:D2_QWC=00000013>
    <VIF1_STAT=00000040:GIF_STAT=10000040>
sceGsPutDrawEnv: DMA Ch.2 does not terminate
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=005dcec0:D1_MADR=005dced0:D1_QWC=00000000>
    <D2_CHCR=00000101:D2_TADR=006ed130:D2_MADR=004be390:D2_QWC=00000013>
    <VIF1_STAT=00000040:GIF_STAT=10000040>
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=005dced0:D1_MADR=005dcee0:D1_QWC=00000000>
    <D2_CHCR=00000101:D2_TADR=006ed130:D2_MADR=004be390:D2_QWC=00000013>
    <VIF1_STAT=00000040:GIF_STAT=10000040>
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=005dced0:D1_MADR=005dcee0:D1_QWC=00000000>
    <D2_CHCR=00000101:D2_TADR=006ed130:D2_MADR=004be390:D2_QWC=00000013>
    <VIF1_STAT=00000040:GIF_STAT=10000040>
libdma: sync timeout
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=004dcec0:D1_MADR=004dced0:D1_QWC=00000000>
    <D2_CHCR=10000105:D2_TADR=006ec240:D2_MADR=006ec210:D2_QWC=00000003>
    <VIF1_STAT=000000c0:GIF_STAT=03000040>
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=004dcec0:D1_MADR=004dced0:D1_QWC=00000000>
    <D2_CHCR=10000105:D2_TADR=006ec240:D2_MADR=006ec210:D2_QWC=00000003>
    <VIF1_STAT=000000c0:GIF_STAT=03000040>
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=004dcec0:D1_MADR=004dced0:D1_QWC=00000000>
    <D2_CHCR=10000105:D2_TADR=006ec240:D2_MADR=006ec210:D2_QWC=00000003>
    <VIF1_STAT=000000c0:GIF_STAT=03000040>
sceGsSyncPath: DMA Ch.2 does not terminate
    <D1_CHCR=70000045:D1_TADR=004dcec0:D1_MADR=004dced0:D1_QWC=00000000>
    <D2_CHCR=10000105:D2_TADR=006ec240:D2_MADR=004be390:D2_QWC=00000013>
    <VIF1_STAT=000000c0:GIF_STAT=10000040>
sceGsPutDrawEnv: DMA Ch.2 does not terminate

It seems to more or less repeat from here. I can put up a full log dump, if
requested.

4) What steps will reproduce the problem?
1. Start the game.
2. Wait for the videos to start.
3. *Sounds of cats yowling, people screaming, Godzilla roaring, etc. etc.
sensational so forth*

5) Please provide any additional information below.

Well, I'm in my mid-twenties, single, and... oh, information related to the
PROBLEM. Well, uh, not much more I can provide that I can think of. I gave
a pretty decent amount of it. Again, if a dev can point me in the direction
of what is the likely culprit, I can at least begin reverting via HEAD
revisions, as testing 411 different revisions will suck very badly. Then
again, maybe my problem points out an immediate, quick fix, and I don't
even need to go through that.

......ladies, feel free to call me anyway.

6) Attachments.  *IMPORTANT!*  We have a limited storage quota on
GoogleCode, so please keep attachments to UNDER 200k!  Do not upload
blockdumps!  Do not upload MP3 dumps.  Use 3rd party image hosts for
screenshots when possible.

Only to this game, I'm afraid. I love it to bits. I'd marry it if I could,
but I can't. Alas...

Original issue reported on code.google.com by evolutionrevolution@gmail.com on 31 May 2010 at 3:44

[GoogleCodeExporter]

One further note, since I can't edit my original post, silly Google Code.

It does take a short while for the video to lock up, about six seconds or so. 
If you
have the game, it should lock up right around when the hourglass is forming.

Original comment by evolutionrevolution@gmail.com on 31 May 2010 at 3:50

[GoogleCodeExporter]

This would be an IPU-related change, most likely.  We've had the DMA Ch.2 
termination 
error in games before, and either rama or Refraction will have a better idea 
what 
causes them (I'm not an IPU expert).

Revisions that modify ipu.cpp or have IPU in the comment log will be likely 
culprits 
for causing this breakage.  Be warned that there's a good chance that the rev 
that 
breaks this game may have fixed many others (so often that happens).

Original comment by Jake.Stine on 1 Jun 2010 at 1:41

• Changed state: Accepted
• Added labels: Component-IPU
• Removed labels: Component-EECore

[GoogleCodeExporter]

Alright. Got my list of culprits. That narrowed it down to 36 revisions. So to 
do
simple math, I went with the direct middle of that, r2879. Busted.

Halve it again. r2829. Works, so our problem's now a bit more narrowed.

r2863. Deader than dead.

r2853. Nixed.

r2830. Working.

r2832. Works.

On a hunch... r2852. Working. DING DING DING.

r2853 is what broke it, and our naughty little member who broke it is 
refraction.

Please hose him down thoroughly, then rub the lotion on its skin anyway.

And, of course, a fix would be appreciated. :P

Original comment by evolutionrevolution@gmail.com on 1 Jun 2010 at 5:50

[GoogleCodeExporter]

Update time. Got a report that using the "EE Timing Hack" Gamefix solves the 
problem,
and indeed it does.

So, the solutions are either a "proper" fix, or this game needes to get added 
to the
database.

If you guys are going to go the database route, let me know (if anything) what 
else I
need to provide so you guys can add it to it.

Original comment by evolutionrevolution@gmail.com on 1 Jun 2010 at 10:08

[GoogleCodeExporter]

looks easy enough to fix, email me on refraction [At] gmail.com  and we will 
sort
this for you :)

Original comment by refraction on 2 Jun 2010 at 1:10

[GoogleCodeExporter]

- Removed Voodoo Cycles << I bet a cold one that it's the same issue that rips 
DDS
videos :p

Original comment by ramapcsx2 on 2 Jun 2010 at 10:26

[GoogleCodeExporter]

you owe me a cold one then ;p the reason the DDS ones rip is due to transfers 
being
half done and ipu trying to take the data xD The FFX video fix sorts that (which
forces IPU to wait for GIF and VIF)

Im betting here DMA gets suspended and because its using gif mfifo (which uses a
different interrupt) it doesnt queue it for resuming and never does ;p

thats my initial guess anyway. why the hack fixes it i dunno :P

Original comment by refraction on 2 Jun 2010 at 10:40

[GoogleCodeExporter]

You'll find out soon enough, refraction. Check your inbox for my message and 
links to
the dumps.

...I still want some Lucky Charms. :(

Original comment by evolutionrevolution@gmail.com on 3 Jun 2010 at 1:06

[GoogleCodeExporter]

Lucky charms are in the post ;p

Original comment by refraction on 3 Jun 2010 at 4:39

[GoogleCodeExporter]

this happens in gungrave ntsc too

Original comment by serban.a...@gmail.com on 8 Jun 2010 at 12:34

[GoogleCodeExporter]

Bloodindark: That might not be the same issue, though, even though it presents 
the same symptoms. If you're getting similar console output, though, it's 
possible.

Anyway, Refraction will likely have this fixed when he gets back (We were 
communicating before his departure) but just in case he hasn't figured it out, 
this might be a hint: The lockup occurs when the movie file ends. The rest of 
that intro is, thus, a non-movie effect.

Original comment by evolutionrevolution@gmail.com on 12 Jun 2010 at 5:01

[GoogleCodeExporter]

well yes i have the same console output but i finished the game like using 
savestatesover and over again ot doing something different in that part of the 
game ehere it hangs like instead of shooting jumping or instead of walking 
forward doing it backward:)) ...i know ..a pain in the ass

Original comment by serban.a...@gmail.com on 12 Jun 2010 at 9:56

[GoogleCodeExporter]

As far as im aware (and able to test) the videos work again, so marking this as 
fixed.

The flashing is due to it being interlaced, it is presuming it is doing 
"alternate line rendering" so its trying to make the "black horizontal line 
effect"

Original comment by refraction on 25 Jun 2010 at 12:50

• Changed state: Fixed

[GoogleCodeExporter]

They did not flicker at all before the changes of r2853. Some interlace modes 
will "fix" it but at the cost of either very obvious interlacing changes, 
darkening the movie, or both. Obviously, in Progressive mode, which the game 
supports, none of these are available.

Using the EE Timing Fix, just like before when they locked the emulator up at 
the end of playback, results in them playing fine with no flicker at all.

Original comment by evolutionrevolution@gmail.com on 25 Jun 2010 at 1:25