The Beta Strikes Back; NuScratch beta release

[timrowledge]

Let's divert discussion from the blog comments to a thread here.

I've copied some points across, feel free to move more.

[timrowledge]

First question -
Q: Can I test it on a different system than my Raspi?
A: You *could* run the image on any recent x86 VM . I run it all the time on an iMac. You need to download a system for your machine from www.squeak.org - there is an all-in-one zip file that has the latest Squeak release for x86 linux, OS X & Windows. Then once you've been able to run a vanilla Squeak you'd need to extract the NuScratch .image file from the Pi and do whatever your OS needs to connect that to be run with the Squeak VM.
If you want to go the other way - run an image file from a Windows or Mac machine on your Pi - then it's as simple as copying the .image & .changes file to your Pi.

[timrowledge]

Thread priority support:-
Spnngman reported - doesn’t work for me: i get:

Code: Select all

pthread_setschedparam failed; consider using ITIMER_HEARTBEAT: Operation not permitted

and Oscar said -
I get

Code: Select all

pthread_setschedparam failed; consider using ITIMER_HEARTBEAT: Operation not permitted

, but when I logged in as root and ran it that didn’t happen.

A: that *probably* implies you're using an older kernel since it is supposed to be ok to change thread priority without having root authority in recent linux kernels.

[simplesi]

This is my first impressions
Fonts don’t look the same as normal - close but not quite the same.

Took much longer to start up than Scratch - 16 seconds of waiting with nothing happening

File open asked me if I wanted to save current project when I hadn’t done anything at all

Also I had to click File then Open then Open again to get it work

Loaded in l larger project - tried moving a set of blocks around - seemed fine

Tried using scissors tool on a wait block to remove it from the middle of set of blocks - nothing happened - CPU load went up to 100% had to restart NuScratch

Works fine with my ScratchGPIO Python handler so the broadcast/sensor updates are working

Simon

[ghp]

Hello,

in some places I have seen some squeak error popup.
Scenario:
have a list with data, export to file system (works)
export again to same file name, there is a error popup

Regards, Gerhard

list_export_twice.png (14.56 KiB) Viewed 7816 times

[ghp]

Hello,

import sounds, Music Loops, Eggs gives a MPEG-Error.

mpeg.png (11.31 KiB) Viewed 7781 times

Sound is 'playable' then, when I choose proceed.

Regards, Gerhard

System: raspbian.

[ghp]

Minor problem:
In 'control' there are the 'when I receive' blocks.
Select one and drag to script area.
open the popup, usually populated with 'new' and available signal names. If you move the mouse over a name or 'new', the entry is highlighted. There is a small gap between the highlighted ares. Drag the popup in the gaps (also possible in the frame around it), and you can move it away. In scratch 1.4, clicking in the gap areas closes the popup.

Regards,
Gerhard

[spongman]

A: that *probably* implies you're using an older kernel since it is supposed to be ok to change thread priority without having root authority in recent linux kernels.

i'm using "Linux raspberrypi 3.10.25+ #622 PREEMPT Fri Jan 3 18:41:00 GMT 2014 armv6l GNU/Linux"

is that too old?

regardless, should this really be a fatal error?

[simplesi]

I'd say - yep - that is 6 months old - need to update - this is new stuff - no point in trying it on old systems

Simon

[ghp]

Hello Tim,

are there plans for a better-beta ? My school team has three weeks vacation now, and I would like to give them the new software on their return because of the better stability in sound. Would be also result in some intensive sessions and test results.
The character set problem (german umlaut), the error popups now and then, the 'letter i from text' conversion and not alpha backgrounds should be fixed till then, all the others I have seen here are lower severity.

What do you think ?

Regards,
Gerhard

[spongman]

I'd say - yep - that is 6 months old - need to update

ok, i've upgraded everything, and i'm on

Code: Select all

Linux raspberrypi 3.12.21+ #688 PREEMPT Wed Jun 4 20:47:24 BST 2014 armv6l GNU/Linux

and i still get the same problem.

again: should it really be a fatal error?

[simplesi]

again: should it really be a fatal error?

Not sure what your asking here - its beta software - if Tim knew where the error was - he could deal with it - he doesn't - so he can't -:)

I wondering if your played around with default user name/passwords? Are you running as pi/raspberry? Has root user password been left at default?
Simon

[timrowledge]

again: should it really be a fatal error?

Not sure what your asking here - its beta software - if Tim knew where the error was - he could deal with it - he doesn't - so he can't -:)

I wondering if your played around with default user name/passwords? Are you running as pi/raspberry? Has root user password been left at default?
Simon

The practical problem is that we either
a) use the thread based time tick which requires a kernel that allows priority changing for threads
b) use the previously tried interrupt based tick that clashes with some deep kernel interrupt issues and ALSA library bugs.

Simon asks a good question about your user settings; for the defaults it seems to work so if we can discern what you might have different maybe we can solve the problem.

One easy test (I hope!) would be to poring with root privileges and try again, and maybe try logging in as normal but running Scratch with sudo?

[timrowledge]

This is my first impressions
Fonts don’t look the same as normal - close but not quite the same.

I've tried to get the fonts as close as possible but fonts are ... fonts. It's almost a swearword. At some point we may be able to make use of a gpu based renderer which would do both better and faster.

Took much longer to start up than Scratch - 16 seconds of waiting with nothing happening

It's all those language files. I'll be taking a look at that next week, along with trying to speed up paint etc opening.

File open asked me if I wanted to save current project when I hadn’t done anything at all

Also I had to click File then Open then Open again to get it work

Tracking the 'dirty' flag is fairly tricky to get exactly right and it's probably a bit enthusiastic right now. Probably better than missing and throwing away a project?

Loaded in l larger project - tried moving a set of blocks around - seemed fine

Good. The block editing seems pretty good right now but I think could be even better at the cost of some changes in behaviour - I'd like to collapse large stacks to a symbolic icon when dragging, for example. I'm sure some people would love it and some would hate it.

Tried using scissors tool on a wait block to remove it from the middle of set of blocks - nothing happened - CPU load went up to 100% had to restart NuScratch

Oops. Looks like you must have sent it out for a long wait. I'd guess there is something hitting an infinite recursion etc.

Works fine with my ScratchGPIO Python handler so the broadcast/sensor updates are working

This is especially welcome news since I don't have any experience with, nor time to tackle, this area. My hope is that nothing will change between old & new for this stuff but if nothing else socket libraries change and could easily screw up life.

[timrowledge]

Scenario:
have a list with data, export to file system (works)
export again to same file name, there is a error popup

Hmm, somewhere the code is expecting a FileDirecotry object and is getting a plain string (probably the filename string). Ought not be too hard to solve, he says glibly...

[timrowledge]

import sounds, Music Loops, Eggs gives a MPEG-Error.

mpeg.png

Sound is 'playable' then, when I choose proceed.

That's a problem closing the file during finalisation of the garbage collector; likely the file handle is futzed somewhere. This might be a pain to fix because finalisation is an asynchronous process and getting a handle on the ... handle... is sometimes difficult to ... handle.

[timrowledge]

In 'control' there are the 'when I receive' blocks.
Select one and drag to script area.
open the popup, usually populated with 'new' and available signal names. If you move the mouse over a name or 'new', the entry is highlighted. There is a small gap between the highlighted ares. Drag the popup in the gaps (also possible in the frame around it), and you can move it away. In scratch 1.4, clicking in the gap areas closes the popup.

That sounds fun. Oh, I see what you mean - it's a slight difference in how the menus are assembled. I can probably clean that up a bit but it is technically intended that you can drag menus out of the way by dragging a not-text place, at least within Squeak. The bigger problem I just spotted is that trying to add more than one new event name removes the previous one. Not quite what is intended, I'm sure!

[timrowledge]

are there plans for a better-beta ?

Oh yes. Very yes.

I've been very occupied with getting the Cog dynamic translation project moving on the ARM for the last few weeks and it requires a fairly strange mind-set and total concentration. In some ways it is fabulous fun but in others it's really hard to sustain. I'm taking a break next week to return to solving some of the problems being reported. The good news is that the Cog is moving along nicely, and should eventually give somewhere between 3-5 x performance improvements (depending heavily on the exact mix of work done) and then we have a new memory model due for release late this year which looks likely to double that, and perhaps next year an adaptive and trainable optimiser that might double or even triple that.

My school team has three weeks vacation now, and I would like to give them the new software on their return because of the better stability in sound.

A whole three weeks? Kids these days get it easy; when I was a kid we had to walk up hill both ways in the snow to get to work at 3am... etc etc.
Yes, with luck a second beta should be around at the end of next week. It should be a touch faster and hopefully fix all the problems reported up to now.

The character set problem (german umlaut), the error popups now and then, the 'letter i from text' conversion and not alpha backgrounds should be fixed till then, all the others I have seen here are lower severity.

Character set issues are probably some problems with unicode handling, which is always such fun.
alpha backgrounds? That doesn't seem familiar?

[timrowledge]

Tried using scissors tool on a wait block to remove it from the middle of set of blocks - nothing happened - CPU load went up to 100% had to restart NuScratch

Oops. Looks like you must have sent it out for a long wait. I'd guess there is something hitting an infinite recursion etc.

Hmm, I can't replicate this yet.
Can you repeat it? If so, a recipe would help track it down.

[ghp]

Scissors-Tool:
I never knew it is possible to use it on scripts, so I was curious to see how it works.

Precondition: assemble a small script of a few simple blocks and C-shapes (while).

in win-scratch-1.4, Click on scissors-tool and move over script. Single blocks get a red border. C-shapes get red border including content. When clicked, the tool cuts the selected blocks. The rest of the script (tail) glides up and closes the gap. For C-shapes, the tail glides up and the c-shape is deleted then.

in RPi-scratch-1.4, there is no red border
cutting blocks works, cutting c-shapes works and shows similar animation as in win-scratch-1.4.

in RPi-scratch-nsb4, there is no red border.
cutting single blocks works either.
cutting c-shapes results in 100% cpu and you need to kill scratch.

Side topic: Tooltip-Text on scissor tool (and its neighbors 'duplicate', 'shrink', 'grow') are missing.
Regards,
Gerhard

[timrowledge]

Scissors-Tool:
I never knew it is possible to use it on scripts, so I was curious to see how it works.

Precondition: assemble a small script of a few simple blocks and C-shapes (while).

I'm still completely unable to replicate this; could you list an exact script that causes the problem? I've tried all the combinations I can come up with, but no problem comes up. Very odd.

And I see what you mean about the red border. I wonder how I lost that...

[timrowledge]

Side topic: Tooltip-Text on scissor tool (and its neighbors 'duplicate', 'shrink', 'grow') are missing.

Hmph. They work on my machine. This is a bit strange.

[timrowledge]

Scenario:
have a list with data, export to file system (works)
export again to same file name, there is a error popup

Hmm, somewhere the code is expecting a FileDirecotry object and is getting a plain string (probably the filename string). Ought not be too hard to solve, he says glibly...

Looks like I've fixed that one at least.

[ghp]

Hello,

you are right, not so simple to reproduce. The following works in neg sense.

scissors.png (1.55 KiB) Viewed 7399 times

The blocks before and after are needed. Tried various combinations of the setup sequence, seems to be independent.

Regards, Gerhard

[timrowledge]

Select one and drag to script area.
open the popup, usually populated with 'new' and available signal names. If you move the mouse over a name or 'new', the entry is highlighted. There is a small gap between the highlighted ares. Drag the popup in the gaps (also possible in the frame around it), and you can move it away. In scratch 1.4, clicking in the gap areas closes the popup.

Fixed.