VMMaker loading uncovers a bug in MC

> I tried to load VMMaker package into latest trunk image:
>
> shortRunLengthAt: i from: runArray
>        "Return the run-length value from the given ShortRunArray."
>        ^((self cCoerce: runArray to:'int *') at: i) bitShift: Argument expected ->- 16
>
>
> I know, this is because of the latest fixes to parser to disallow
> space after unary minus.
>
> Once i fixed that and accepted the change, next thing which i met is DNU,
> because MC tries to install nil as a method into a class.
>
> It seems that a method which is hot-fixed is lost somewhere and
> instead, somehow MC includes nil in its methodAdditions collection :(
>
> Also, before that this happened with another package i tried to load,
> which containing syntax errors while loading.
> So, this error is easy to reproduce.
>
> --
> Best regards,
> Igor Stasenko AKA sig.
>

> On Mon, Mar 08, 2010 at 03:09:36AM +0200, Igor Stasenko wrote:
>> I tried to load VMMaker package into latest trunk image:
>>
>> shortRunLengthAt: i from: runArray
>>       "Return the run-length value from the given ShortRunArray."
>>       ^((self cCoerce: runArray to:'int *') at: i) bitShift: Argument expected ->- 16
>>
>>
>> I know, this is because of the latest fixes to parser to disallow
>> space after unary minus.
>
> This was fixed in VMMaker-dtl.159.
>
>> ohh sorry for not mentioning. I used 'update-dtl.2.mcm' config for loading it.
>
> So I guess I need to fix the configuration map. I updated it to
> 'update-dtl.3.mcm' which hopefully takes care of the VMMaker part
> of the problem.
>

> On 8 March 2010 05:00, David T. Lewis <[hidden email]> wrote:
> > On Mon, Mar 08, 2010 at 03:09:36AM +0200, Igor Stasenko wrote:
> >> I tried to load VMMaker package into latest trunk image:
> >>
> >> shortRunLengthAt: i from: runArray
> >> ?? ?? ?? "Return the run-length value from the given ShortRunArray."
> >> ?? ?? ?? ^((self cCoerce: runArray to:'int *') at: i) bitShift: Argument expected ->- 16
> >>
> >>
> >> I know, this is because of the latest fixes to parser to disallow
> >> space after unary minus.
> >
> > This was fixed in VMMaker-dtl.159.
> >
> >> ohh sorry for not mentioning. I used 'update-dtl.2.mcm' config for loading it.
> >
> > So I guess I need to fix the configuration map. I updated it to
> > 'update-dtl.3.mcm' which hopefully takes care of the VMMaker part
> > of the problem.
> >
> Everything is ok now, except that once it gets to loading a freetype,
> i got 'subscript out of bounds'
> and image UI become unusable, since same error happens over and over
> again when rendering text.
>

> On Mon, Mar 08, 2010 at 05:14:05AM +0200, Igor Stasenko wrote:
>> On 8 March 2010 05:00, David T. Lewis <[hidden email]> wrote:
>>> On Mon, Mar 08, 2010 at 03:09:36AM +0200, Igor Stasenko wrote:
>>>> I tried to load VMMaker package into latest trunk image:
>>>>
>>>> shortRunLengthAt: i from: runArray
>>>> ?? ?? ?? "Return the run-length value from the given ShortRunArray."
>>>> ?? ?? ?? ^((self cCoerce: runArray to:'int *') at: i) bitShift: Argument expected ->- 16
>>>>
>>>>
>>>> I know, this is because of the latest fixes to parser to disallow
>>>> space after unary minus.
>>>
>>> This was fixed in VMMaker-dtl.159.
>>>
>>>> ohh sorry for not mentioning. I used 'update-dtl.2.mcm' config for loading it.
>>>
>>> So I guess I need to fix the configuration map. I updated it to
>>> 'update-dtl.3.mcm' which hopefully takes care of the VMMaker part
>>> of the problem.
>>>
>> Everything is ok now, except that once it gets to loading a freetype,
>> i got 'subscript out of bounds'
>> and image UI become unusable, since same error happens over and over
>> again when rendering text.
>>
>
> Eeek! I just tried loading from http://www.squeaksource.com/FreeTypePlus
> and something is definitely broken just as you describe.
>
> The error is in WorldState>>displayWorldSafely leading to an Array
> subscript out of bounds. It's not possible to open a debugger.

> On Sun, 7 Mar 2010, David T. Lewis wrote:
>
>> On Mon, Mar 08, 2010 at 05:14:05AM +0200, Igor Stasenko wrote:
>>>
>>> On 8 March 2010 05:00, David T. Lewis <[hidden email]> wrote:
>>>>
>>>> On Mon, Mar 08, 2010 at 03:09:36AM +0200, Igor Stasenko wrote:
>>>>>
>>>>> I tried to load VMMaker package into latest trunk image:
>>>>>
>>>>> shortRunLengthAt: i from: runArray
>>>>> ?? ?? ?? "Return the run-length value from the given ShortRunArray."
>>>>> ?? ?? ?? ^((self cCoerce: runArray to:'int *') at: i) bitShift:
>>>>> Argument expected ->- 16
>>>>>
>>>>>
>>>>> I know, this is because of the latest fixes to parser to disallow
>>>>> space after unary minus.
>>>>
>>>> This was fixed in VMMaker-dtl.159.
>>>>
>>>>> ohh sorry for not mentioning. I used 'update-dtl.2.mcm' config for
>>>>> loading it.
>>>>
>>>> So I guess I need to fix the configuration map. I updated it to
>>>> 'update-dtl.3.mcm' which hopefully takes care of the VMMaker part
>>>> of the problem.
>>>>
>>> Everything is ok now, except that once it gets to loading a freetype,
>>> i got 'subscript out of bounds'
>>> and image UI become unusable, since same error happens over and over
>>> again when rendering text.
>>>
>>
>> Eeek! I just tried loading from http://www.squeaksource.com/FreeTypePlus
>> and something is definitely broken just as you describe.
>>
>> The error is in WorldState>>displayWorldSafely leading to an Array
>> subscript out of bounds. It's not possible to open a debugger.
>
> I uploaded two packages to the Inbox which (seem to be) fixing the issue.
> After evaluating this in an updated Trunk image (9634) FreeType should load
> normally:
>
> (Installer squeak project: 'inbox')
>        install: 'Graphics-ul.117.mcz';
>        install: 'Multilingual-ul.100.mcz'.
> CharacterScanner initialize.
>
>
> Levente
>
> P.S. I uploaded these to the Inbox because I'm not familiar with this part
> of Squeak at all, and there's a bit cleaner way to implement these fixes if
> they are correct.
>
>>
>> Dave
>>
>>
>>
>
>

> I uploaded two packages to the Inbox which (seem to be) fixing the
> issue. After evaluating this in an updated Trunk image (9634) FreeType
> should load normally:
>
> (Installer squeak project: 'inbox')
> install: 'Graphics-ul.117.mcz';
> install: 'Multilingual-ul.100.mcz'.
> CharacterScanner initialize.
>
>
> Levente
>
> P.S. I uploaded these to the Inbox because I'm not familiar with this
> part of Squeak at all, and there's a bit cleaner way to implement these
> fixes if they are correct.

> On 3/7/2010 8:06 PM, Levente Uzonyi wrote:
>>
>> I uploaded two packages to the Inbox which (seem to be) fixing the
>> issue. After evaluating this in an updated Trunk image (9634) FreeType
>> should load normally:
>>
>> (Installer squeak project: 'inbox')
>> install: 'Graphics-ul.117.mcz';
>> install: 'Multilingual-ul.100.mcz'.
>> CharacterScanner initialize.
>>
>>
>> Levente
>>
>> P.S. I uploaded these to the Inbox because I'm not familiar with this
>> part of Squeak at all, and there's a bit cleaner way to implement these
>> fixes if they are correct.
>
> Nicolas should look at the changes. I think this is a side effect of the
> Unicode leading char updates he posted.
>
> Cheers,
>  - Andreas
>
>

> 2010/3/8 Andreas Raab <[hidden email]>:
>> On 3/7/2010 8:06 PM, Levente Uzonyi wrote:
>>>
>>> I uploaded two packages to the Inbox which (seem to be) fixing the
>>> issue. After evaluating this in an updated Trunk image (9634) FreeType
>>> should load normally:
>>>
>>> (Installer squeak project: 'inbox')
>>> install: 'Graphics-ul.117.mcz';
>>> install: 'Multilingual-ul.100.mcz'.
>>> CharacterScanner initialize.
>>>
>>>
>>> Levente
>>>
>>> P.S. I uploaded these to the Inbox because I'm not familiar with this
>>> part of Squeak at all, and there's a bit cleaner way to implement these
>>> fixes if they are correct.
>>
>> Nicolas should look at the changes. I think this is a side effect of the
>> Unicode leading char updates he posted.
>>
>> Cheers,
>>  - Andreas
>>
>>
>
> Yes,we cross-posted.
> Take Levente workaround, it's just perfect.
>
> Nicolas
>

> On Mon, 8 Mar 2010, Nicolas Cellier wrote:
>
>> OK, I re-analyzed the stop conditions to sort the mess out.
>> Take for example
>> MultiCharacterScanner >> scanMultiCharactersFrom: startIndex to:
>> stopIndex in: sourceString rightX: rightX stopConditions: stops kern:
>> kernDelta
>>
>> Once upon a time, the test was:
>>                (encoding = 0 and: [ascii < stopConditions size and:
>> [(stopConditions at: ascii + 1) ~~ nil]]) ifTrue: [^ stops at: ascii +
>> 1].
>> There was a conflict with endOfRun/CrossedX in case of ascii=256 or
>> ascii=257 (A macron...)
>> But this did not happen with Unicode leadingChar = 15 (Unicode).
>>
>> This code was the starting point for reifying a TextStopConditions object.
>>
>> Then Andreas tested with 256 instead of stopConditions size:
>>                (encoding = 0 and: [ascii < 256 and: [(stops at: ascii + 1)
>> notNil]])
>>                        ifTrue: [^ stops at: ascii + 1].
>> This was a more lightweight solution, no TextStopConditions object
>> required anymore.
>> The A macron potential bug was out.
>>
>> Then FreeType was merged, reintroducing
>>                (encoding = 0 and: [ascii < stopConditions size and:
>> [(stopConditions at: ascii + 1) ~~ nil]]) ifTrue: [^ stops at: ascii +
>> 1].
>> The A macron potential bug was back in.
>>
>> Then I corrected, because the instance variable stopConditions was
>> shadowing the argument stops:
>>                (encoding = 0 and: [ascii < stops size and: [(stops at:
>> ascii + 1)
>> ~~ nil]]) ifTrue: [^ stops at: ascii + 1].
>> The A macron potential bug was still in.
>>
>> So I wonder if I will not revert to the simple stopConditions
>> class=Array solution with Andreas test ascii < 256.
>> The main reason is for efficiency, this is a critical internal
>> CPU-consuming loop.
>
> I think we should keep TextStopCondition, but it should be a variable
> subclass. This gives us fast accessing, it can have 256 or 258 slots (for
> backwards compatiblity, if necessary) and have the instance variables and
> accessors for endOfRun and crossedX.
>
>
> Levente
>

>> Unless we want to handle control characters with codePoint >= 256 ?
>>
>> What do you think ?
>>
>> Nicolas
>>
>> 2010/3/8 Nicolas Cellier <[hidden email]>:
>>>
>>> 2010/3/8 Andreas Raab <[hidden email]>:
>>>>
>>>> On 3/7/2010 8:06 PM, Levente Uzonyi wrote:
>>>>>
>>>>> I uploaded two packages to the Inbox which (seem to be) fixing the
>>>>> issue. After evaluating this in an updated Trunk image (9634) FreeType
>>>>> should load normally:
>>>>>
>>>>> (Installer squeak project: 'inbox')
>>>>> install: 'Graphics-ul.117.mcz';
>>>>> install: 'Multilingual-ul.100.mcz'.
>>>>> CharacterScanner initialize.
>>>>>
>>>>>
>>>>> Levente
>>>>>
>>>>> P.S. I uploaded these to the Inbox because I'm not familiar with this
>>>>> part of Squeak at all, and there's a bit cleaner way to implement these
>>>>> fixes if they are correct.
>>>>
>>>> Nicolas should look at the changes. I think this is a side effect of the
>>>> Unicode leading char updates he posted.
>>>>
>>>> Cheers,
>>>>  - Andreas
>>>>
>>>>
>>>
>>> Yes,we cross-posted.
>>> Take Levente workaround, it's just perfect.
>>>
>>> Nicolas
>>>
>>
>
>
>
>

> On Mon, Mar 08, 2010 at 04:03:00PM +0200, Igor Stasenko wrote:
>> Guys, can anyone explain me, what i should load to get a working set
>> of VMMaker/plugins packages in trunk image?
>>
>> I'm not interested right now in digging these bugs, since i focused on
>> making changes to VMMaker (i promised that),
>> and bring that to community to disscuss the pros and cons.
>
> If you load just the latest VMMaker itself from SqueakSource, you will
> have enough for a workable VM(*). The MCM update ('VMMaker updateFromServer')
> brings in a few more commonly used plugins from other sources (but not
> all of the plugins distributed on all platforms).
>

> On Mon, 8 Mar 2010, Igor Stasenko wrote:
>
>> Guys, can anyone explain me, what i should load to get a working set
>> of VMMaker/plugins packages in trunk image?
>>
>> I'm not interested right now in digging these bugs, since i focused on
>> making changes to VMMaker (i promised that),
>> and bring that to community to disscuss the pros and cons.
>
> Just update your Trunk image and you can load VMMaker from the mcm.
>

>
> Levente
>
>>
>>
>> --
>> Best regards,
>> Igor Stasenko AKA sig.
>>
>>
>
>

> On 8 March 2010 16:50, Levente Uzonyi <[hidden email]> wrote:
>> On Mon, 8 Mar 2010, Igor Stasenko wrote:
>>
>>> Guys, can anyone explain me, what i should load to get a working set
>>> of VMMaker/plugins packages in trunk image?
>>>
>>> I'm not interested right now in digging these bugs, since i focused on
>>> making changes to VMMaker (i promised that),
>>> and bring that to community to disscuss the pros and cons.
>>
>> Just update your Trunk image and you can load VMMaker from the mcm.
>>
>
> I tried that - updated the image and then tried to load mcm from
> vmmaker repo. And met same problems.

>
> Then i manually loaded FFI, VMMaker
> and then did VMMaker updateFromServer, and it worked fine.
>
>>
>> Levente
>>
>>>
>>>
>>> --
>>> Best regards,
>>> Igor Stasenko AKA sig.
>>>
>>>
>>
>>
>
>
>
> --
> Best regards,
> Igor Stasenko AKA sig.
>
>