refresh problem with etoys in joyride-2301 (8.2.0 update 2)

[mzimmerm]

Environment: I ran olpc-update joyride-2301 I updated al activities to latest from control panel

when running etoys, and put cursor on top of the car, some bits do not get refreshed. there ar othr refresh problems, I think all related to the mouse cursor image having non-transparent background.

[bert]

Yes we know, this is caused by #7612.

A temporary work-around is to disable usage of X shared memory by removing the -xshm option from /usr/bin/etoys (causing a performance degradation but makes the artefacts go away).

[mzimmerm]

i see sorry for the duplicate and thanks for the workaround.

[bert]

No need to apologize - it's goof to have a ticket in the etoys component, too.

[bert]

hehe, I meant "good"

[bert]

We need to decide what to do for the 8.2 release.

A different work-around is disabling the big cursor, but the tiny one is really hardly visible on the XO's screen.

[bert]

Jordan writes in #7612 that we should better go ahead and apply a workaround.

Adding a work-around is trivial, so 8.2.0:+, but which one do we want? I'd rather disable Xshm than having a tiny mouse pointer.

[ohshima]

Replying to bert:

Adding a work-around is trivial, so 8.2.0:+, but which one do we want? I'd rather disable Xshm than having a tiny mouse pointer.

Alternatively, we could have an aliased bigger cursor (assuming that it works).

[bert]

Replying to ohshima:

Alternatively, we could have an aliased bigger cursor (assuming that it works).

I just tried that, I can confirm a 32x32 1-bit cursor works without artifacts even if Xshm is enabled. But we would need to add yet another support code function to the VM (ioSetCursorWithMaskBig?) and more in-image support ... Are you going to code that up? Disabling Xshm on the XO only is a two-line patch.

[cscott]

I'm going to disable Xshm on the 8.2 stream, but leave it enabled in joyride in the hopes that we can actually fix the root cause.

[bert]

Workaround tested working in 8.2-754.

[mzimmerm]

Replying to bert:

Workaround tested working in 8.2-754.

it worked in 754, but it is broken again in 8.2-756

[bert]

Replying to cscott:

I'm going to disable Xshm on the 8.2 stream, but leave it enabled in joyride in the hopes that we can actually fix the root cause.

Since you re-enabled Xshm globally, could you please disable it for Etoys specifically in the 8.2 stream, by deleting the -xshm option in /usr/bin/etoys? Thanks!

[bert]

Replying to bert:

Replying to cscott:

I'm going to disable Xshm on the 8.2 stream, but leave it enabled in joyride in the hopes that we can actually fix the root cause.

Since you re-enabled Xshm globally, could you please disable it for Etoys specifically in the 8.2 stream, by deleting the -xshm option in /usr/bin/etoys? Thanks!

I have a better workaround without a performance penalty, as suggested by Yoshiki above. The next /usr/bin/etoys script will have a line like this:

export SQUEAK_FAKEBIGCURSOR=0

Please change that to 1 to work around the bug. It will also require a new Squeak VM package that I am about to release.

[bert]

So the new packages are in 8.2-757, now pilgrim needs to enable the workaround. Or do you want me to enable it by default in a new package?

[cscott]

Pilgrim fix checked in for next 8.2 build.

[dsd]

Fixed properly in xorg-x11-server-1.4.99.906-2.olpc3.2, coming soon to a joyride near you. We should not need the workaround

[bert]

Replying to dsd:

Fixed properly in xorg-x11-server-1.4.99.906-2.olpc3.2

Fantastic!

[mzimmerm]

Thanks - I will wait for 8.2-758 to appear and test it.

[bert]

works - please remove enanling the workaround.

[dsd]

Let's track this in #7612