Compiling CM9 (and maybe CM10) for Blade

[Guest shezar]

Something's telling me it's not that commit that's causing it. ( https://github.com/C...7a7d1f7f75b2eb2 )

Simply because it was comitted 24 days ago... It's too far in the past and we definitelly had smooth UI in builds after that time.

If you look at it, it got merged much later (https://github.com/CyanogenMod/android_frameworks_base/commit/add0200d0732f70918bf89520ff394d1cf82dac1#diff-15)

[Guest WildeRNS]

kernel/zte/zte-kernel-msm7x27

thank you!

apply as patch or cherry-pick?

[Guest shezar]

thank you!

apply as patch or cherry-pick?

I did it as a patch but you'll need to do it again after each repo sync until it gets merged.

[Guest WildeRNS]

I did it as a patch but you'll need to do it again after each repo sync until it gets merged.

ok, I know that, thank you

[Guest shmizan]

So you are saying that before 13th of May, CM9 was as fast/smooth as ColdfusionX?

yes. the differences were really minor. the clock opened slowly. the app drawer too, and general menu transitions were just a bit slower.

this gerrit change should have made it square. but it wasn't all. that is what tilal and sej are talking about here. apparently they don't suffer the same lag we do.

[Guest C3C0]

If you look at it, it got merged much later (https://github.com/C...f82dac1#diff-15)

Right. I was just looking at it.

[Guest shmizan]

C3C0 you were saying something about testing a build that reverts this change? any result?

[Guest C3C0]

No. Unfotunatelly, revert breaks the build. There must be some dependand commit. Or I am doing something wrong...

[Guest shezar]

No. Unfotunatelly, revert breaks the build. There must be some dependand commit. Or I am doing something wrong...

There must be because if it worked before, decideTextureTarget must have been defined somewhere. That's what my compiler is missing.

[Guest targetbsp]

eh, what lag is that now then?

The issue you noted as causing the clock to look glitchy. On my phone for certain apps it's most noticeable as the animation for apps opening and closing momentarily lagging mid way through and looking very weird. There was that whole true/false thing which fixed that.

Shezar was testing whether that patch was the cause of the other slow downs and it wasn't. So I was expressing my relief that that specifc patch won't need reverted.

Beyond that... I have entirely lost track of this discussion. :D

[Guest jventura]

Beyond that... I have entirely lost track of this discussion. :D

We are talking about this lag:

(...) it seems sej and targetbsp both don't get what shezar, jventura and I get, which is COMPLETE lag in launcher, status bar, windows transition, everything.

[Guest shezar]

We are talking about this lag:

Indeed, with C3C0 we were looking at possible causes for the slowdown and we found a commit which could be responsible. The problem is that we could not revert it to see if that is the case because that broke the build.

Edited May 18, 2012 by shezar

[Guest targetbsp]

All I can say with certainty is that this patch fixed an animation lag problem for me: http://review.cyanogenmod.com/#/c/16056/

I wouldn't like to say I don't see the other lag you guys are seeing. I haven't noticed anything out of the ordinary but I'm generally running CFX and only tend to run CM9 to test pending patches for Talil. I ran CM9 yesterday afternoon and evening and didn't notice anything out of the ordinary.

But I'm not using an ICS based launcher - I'm still using ADW.Ex. And it's laggy anyway! I just haven't got around to finding an ICS launcher I like yet.

So don't take the fact that I haven't seen an issue yet as meaning anything! I'll try and get CM9 on later and have a play with an ICS launcher and the status bar.

[Guest shmizan]

the lag cannot be missed. it's obvious as it makes the ROM unusable.

Indeed, with C3C0 we were looking at possible causes for the slowdown and we found a commit which could be responsible. The problem is that we could not revert it to see if that is the case because that broke the build.

did you try applying this and this? or just one of them?

Edited May 18, 2012 by shmizan

[Guest shezar]

did you try applying this and this? or just one of them?

You're my man! I couldn't find the libqcomui revert so that's why it failed. Thanks!

[Guest shmizan]

please report if it solved the problem or not

[Guest shezar]

please report if it solved the problem or not

I'm building, it takes 40 minutes + a little. I'll edit this one with the results. But I'm quite confident that this solves our problem.

Edit: compile failed on Settings app, lol. Maybe I didn't clean before build, I'll try again.

jventure: a clean build takes 6 hours for me :)

Sorry, I'll do a clean build at night, I need my PC right now. But I'm still confident that these reverts solve our problem.

Edited May 18, 2012 by shezar

[Guest jventura]

I'm building, it takes 40 minutes + a little. I'll edit this one with the results. But I'm quite confident that this solves our problem.

I'm building it also with the reverts, but it will take more than 40 minutes for me.. :)

[Guest sej7278]

just flashing a clean build with the 2 reverts (plus blackscreen and common_mini_phone.mk)

i'll upload it if its a goodun'

[Guest shmizan]

awesomeness. by the way you didn't see the big lag we were talking about? you just trust us that it's there and revert those 2 changes?

oh and what about tilal's 2 line changes to make the clock squish and other stuff equal on CFX and CM9? we dropped that?

[Guest sej7278]

awesomeness. by the way you didn't see the big lag we were talking about? you just trust us that it's there and revert those 2 changes?

oh and what about tilal's 2 line changes to make the clock squish and other stuff equal on CFX and CM9? we dropped that?

yeah the 2 reverts seem to have increased the speed of things, certainly app drawer seems better.

the only lag i saw before was in the notification power widget, and pulling down the notification bar, they're instant now. i didn't see a problem between windows - other than an initial pause rendering the page, but it wasn't jerky or anything - was that what you meant?

i've been going on about CM9 being slower than CFX for so long i think i've got used to it being slower, and i've not tried CFX in a month.

the clock squish is better as i left tilal's 2 lines alone (they are removed, so get true/false by default). it still squishes the screen but it doesn't lag when doing it and look pixelated. i'm starting to think that squishing is a feature of clock, not a bug lol.

so i wonder if CM9 is as fast as CFX now, or if we still need to look into using the source-built GLES lib rather than the prebuilt binary.....?

its uploaded if you're interested, otherwise happy building! (note: no video recording patches in this build).

Edited May 18, 2012 by sej7278

[Guest shmizan]

why wasting your bandwidth when I can waste my cpu ;) let's hope they solve this as soon as possible. the screen fix needs to get merged too. Tom must be still working on it.

I really want to get to the bottom of the keyboard lag too. how can I make a log of it? it only happens when the keyboard needs to fix words in a reasonable speed. it just get stuck and the keyboard drops only to come back 10 seconds later.

edit: doesn't happen in English. happens with Hebrew.

Edited May 18, 2012 by shmizan

[Guest shezar]

Thank you, Simon, I see you commented on Gerrit, too. It's funny to see that you considered that choppy UI smooth :P Thanks for the build, I'll install it but I'm still gonna make my own later with more patches :P

What method do you recommend to apply gerrit patches? Sometimes patch wants to overwrite previous patches so I'm unable to do it right. Could you explain it a bit or point me to a direction with some info about it?

[Guest sej7278]

Thank you, Simon, I see you commented on Gerrit, too. It's funny to see that you considered that choppy UI smooth :P Thanks for the build, I'll install it but I'm still gonna make my own later with more patches :P

What method do you recommend to apply gerrit patches? Sometimes patch wants to overwrite previous patches so I'm unable to do it right. Could you explain it a bit or point me to a direction with some info about it?

i usually just make a patch file from gerrit's patch-anon-http command, e.g.

git fetch http://review.cyanogenmod.com/CyanogenMod/android_frameworks_base refs/changes/28/16028/1 && git format-patch -1 --stdout FETCH_HEAD > ~/frameworks_revert.patch

then i can reset my local repo to HEAD so repo sync won't throw a wobbly:

git reset --hard HEAD[/code]




then i apply the patch again (from the correct directory):

[CODE]git apply ~/frameworks_revert.patch[/code]

i don't bother messing around making branches and stuff unless i'm likely to be uploading fixes.

i think tom_g said he wanted to look into the backlight patch a bit more as what he was doing was hacky or something. i still don't see why he's not convinced about common_mini_phone.mk though.......

edit: doesn't happen in English. happens with Hebrew.

there's a joke there somewhere, but i'm not going anywhere near it on a Friday :D

Edited May 18, 2012 by sej7278

[Guest shezar]

Whoaaa, now this is smooooooth :)

Thanks for the build and the info, it was most helpful.

Also I'm guessing TomG wants to have a build as full as possible.

Edited May 18, 2012 by shezar