Smalltalk › Pharo › Pharo Smalltalk Developers

Pharo 2.0 Jenkins build failing

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

12 messages Options

Sean P. DeNigris

Pharo 2.0 Jenkins build failing

Administrator

It looks like it's related to the progress bar overhaul... I have a pretty good idea what's going on... I'm working on it now...

Sean

Cheers,
Sean

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

I think I have a fix... The bug was the new progress bar wasn't integrated with headless mode... just testing now and I'll upload...

- S

Cheers,
Sean

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

Fix attached to:
Issue 6055: [BUG] New System Progress doesn't work when headless
http://code.google.com/p/pharo/issues/detail?id=6055

2.0

When headless and displaying progress, with used to be a no-op, now there is a BlockClosure DNU label:. This is because progress used to use a block that was pretending to be a class, so #value: would get sent everywhere, with the "selector" as the argument (e.g. value: #label). Now that we have a real progress object, we are sending #label: explicitly. Since CommandLineUIManager wasn't updated, it's still passing a block ([:bar |]), which obviously doesn't understand #label:

The attached st files fix CommandLineUIManager to use a DummySystemProgressItem instead of a block.

There will be more refactoring of the entire system progress framework, but let's get it working, one step at a time.

N.B. The DummySystemProgressItem was temporarily placed in its own category to avoid making Morphic-Widgets dirty, which would be wiped out as the updates to 20119 are loaded

Cheers,
Sean

Stéphane Ducasse

Re: Pharo 2.0 Jenkins build failing

Thanks Sean.
I'm going to paris for a presentation so I may have a look tomorrow evening.

Stef

On Jun 11, 2012, at 6:33 PM, Sean P. DeNigris wrote:

> Fix attached to:
> Issue 6055: [BUG] New System Progress doesn't work when headless
> http://code.google.com/p/pharo/issues/detail?id=6055
>
>
>
>> 2.0
>>
>> When headless and displaying progress, with used to be a no-op, now there
>> is a BlockClosure DNU label:. This is because progress used to use a block
>> that was pretending to be a class, so #value: would get sent everywhere,
>> with the "selector" as the argument (e.g. value: #label). Now that we have
>> a real progress object, we are sending #label: explicitly. Since
>> CommandLineUIManager wasn't updated, it's still passing a block ([:bar
>> |]), which obviously doesn't understand #label:
>>
>> The attached st files fix CommandLineUIManager to use a
>> DummySystemProgressItem instead of a block.
>>
>> There will be more refactoring of the entire system progress framework,
>> but let's get it working, one step at a time.
>>
>> N.B. The DummySystemProgressItem was temporarily placed in its own
>> category to avoid making Morphic-Widgets dirty, which would be wiped out
>> as the updates to 20119 are loaded
>>
>
> --
> View this message in context: http://forum.world.st/Pharo-2-0-Jenkins-build-failing-tp4634306p4634339.html
> Sent from the Pharo Smalltalk mailing list archive at Nabble.com.
>

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

Stéphane Ducasse wrote

Thanks Sean.
I'm going to paris for a presentation so I may have a look tomorrow evening.

No problem. If anybody needs the latest update (#20129 currently) before then, the st files attached to the issue are only needed to update heedlessly.

Otherwise, you can download 20119 from Jenkins and do a "software update" via the world menu (or programmatically).

n.b. the latest Cog Jit VM on Jenkins crashes when doing this update. The latest Jenkins Stack VM works fine.

HTH,
Sean

Cheers,
Sean

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

Sean P. DeNigris wrote

the latest Cog Jit VM on Jenkins crashes when doing this update. The latest Jenkins Stack VM works fine.

Disregard that. OS X was picking up an old vm on my system. The latest Jit and Stack VMs both work.

Cheers,
Sean

Pavel Krivanek-3

Re: Pharo 2.0 Jenkins build failing

In reply to this post by Sean P. DeNigris

But will it work on Jenkins? Because the update process will want to
process the updates one by one so Jenkins will want to load this
patches too late (as part of later update). I guess we will need to
revert the last updates.

-- Pavel

On Tue, Jun 12, 2012 at 3:20 AM, Sean P. DeNigris <[hidden email]> wrote:

>
> Stéphane Ducasse wrote
>>
>> Thanks Sean.
>> I'm going to paris for a presentation so I may have a look tomorrow
>> evening.
>>
>
> No problem. If anybody needs the latest update (#20129 currently) before
> then, the st files attached to the issue are only needed to update
> heedlessly.
>
> Otherwise, you can download 20119 from Jenkins and do a "software update"
> via the world menu (or programmatically).
>
> n.b. the latest Cog Jit VM on Jenkins crashes when doing this update. The
> latest Jenkins Stack VM works fine.
>
> HTH,
> Sean
>
> --
> View this message in context: http://forum.world.st/Pharo-2-0-Jenkins-build-failing-tp4634306p4634400.html
> Sent from the Pharo Smalltalk mailing list archive at Nabble.com.
>

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

Pavel Krivanek-3 wrote

But will it work on Jenkins? Because the update process will want to
process the updates one by one so Jenkins will want to load this
patches too late (as part of later update). I guess we will need to
revert the last updates.

Hmm... it seemed like a *lot* of work for Steph to get the FS-everywhere and progress bar updates in... is there no way to sneak the st files in just to get it working? Can we temporarily change the build script to insert the sts or maybe just wrap it in an error handler that catches invalid dates coming out of the zip reading method...

Once we're up to latest update, I can just make a slice...

Sean

Cheers,
Sean

Pavel Krivanek-3

Re: Pharo 2.0 Jenkins build failing

On Tue, Jun 12, 2012 at 3:53 PM, Sean P. DeNigris <[hidden email]> wrote:

>
> Pavel Krivanek-3 wrote
>>
>> But will it work on Jenkins? Because the update process will want to
>> process the updates one by one so Jenkins will want to load this
>> patches too late (as part of later update). I guess we will need to
>> revert the last updates.
>>
>
> Hmm... it seemed like a *lot* of work for Steph to get the FS-everywhere and
> progress bar updates in... is there no way to sneak the st files in just to
> get it working? Can we temporarily change the build script to insert the sts
> or maybe just wrap it in an error handler that catches invalid dates coming
> out of the zip reading method...
>
> Once we're up to latest update, I can just make a slice...
>
> Sean

I hope it will be possilbe to create packages with the progress bar
patches and then to create one all-embracing update simply using the
latest packages versions.

-- Pavel

>
> --
> View this message in context: http://forum.world.st/Pharo-2-0-Jenkins-build-failing-tp4634306p4634477.html
> Sent from the Pharo Smalltalk mailing list archive at Nabble.com.
>

Stéphane Ducasse

Re: Pharo 2.0 Jenkins build failing

What is the problem?
Because with camillo we still have some change to get in such as the compression changes and monticello for FS.

Now for the progress bar the best way is that we disable it, load your change and reenable it.

Are you changes loading?
Then can we load something after :)

Are the two most important questions :)

Stef

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

Stéphane Ducasse wrote

What is the problem?
Are you changes loading?
Then can we load something after :)

The problem is actually just before the new progress bar gets installed, in that same update (#20120). A related change to Collections-Abstract (Collection>>do:displayingProgress:every:) is causing a DNU when headless. I think I have a solution. I emailed you and esteban...

Cheers,
Sean

Sean P. DeNigris

Re: Pharo 2.0 Jenkins build failing

Administrator

The build is now green! The last successful build is update #20132. Thank you Steph, Esteban, and Marcus :)

In the future, if anyone wants to update 2.0 headlessly past #20119, you will need to wrap the update code in an error handler like the following (headful updating will work out of the box without this):
[ UpdateStreamer new beSilent; updateFromServer ]
on: MessageNotUnderstood
do: [ :e |
(e message selector = #label: and: [ e receiver isKindOf: BlockClosure]) ifTrue: [ e resume ] ]

Cheers,
Sean

Cheers,
Sean