Smalltalk › Squeak › Squeak - Dev

More problems with Monticello 447

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

5 messages Options

Nicolas Cellier

Re: More problems with Monticello 447

2011/3/26 Nicolas Cellier <[hidden email]>:
> One of my images crashed on an infinite loop.
> Don't know the origin of the problem, but I had to change
>
> MCVersionName>>hash
> ^ (String withAll: self versionName) hash
>

I still think (self versionName hash) is very dangerous

> I also have CurrentReadOnlySourceFiles undefined, but maybe I missed
> some preamble/postscript ?
>

Never mind, the second was a load order problem (I had some local
changes in MC so I did a manual update).

> Nicolas
>

Levente Uzonyi-2

Re: More problems with Monticello 447

In reply to this post by Nicolas Cellier

On Sat, 26 Mar 2011, Nicolas Cellier wrote:

> One of my images crashed on an infinite loop.
> Don't know the origin of the problem, but I had to change
>
> MCVersionName>>hash
> ^ (String withAll: self versionName) hash

#versionName always returns a ByteString, so it won't be recursive.
Because of this, it's not worth putting MCVersionName to hashed
collections, because the performance will be poor.

>
> I also have CurrentReadOnlySourceFiles undefined, but maybe I missed
> some preamble/postscript ?

There's no preamble/postscript related to CurrentReadOnlySourceFiles, it's
in Files-ul.104. There's also an mcm to ensure the proper load order, so I
guess it's only a problem with your image.

Levente

>
> Nicolas
>
>

Nicolas Cellier

Re: More problems with Monticello 447

2011/3/26 Levente Uzonyi <[hidden email]>:

> On Sat, 26 Mar 2011, Nicolas Cellier wrote:
>
>> One of my images crashed on an infinite loop.
>> Don't know the origin of the problem, but I had to change
>>
>> MCVersionName>>hash
>> ^ (String withAll: self versionName) hash
>
> #versionName always returns a ByteString, so it won't be recursive. Because
> of this, it's not worth putting MCVersionName to hashed collections, because
> the performance will be poor.
>

OK, don't know what went on, I probably messed with load order or something.
There was a removed (String>>versionName ^self) that might explain...

However, I still consider this construction as fragile, since it
relies on species.
A guiding comment in MCVersionName>>species would be welcome.
It would be even nicer to assert the implementation with a TestCase like:

self assert: MCVersionName someInstance species ~= MCVersionName.

Since #first: might be changed in the future and not use #species
anymore (who knows...) ,
then it might be even better to document the intention with a TestCase like:

self assert: MCVersionName someInstance versionName class ~= MCVersionName.

Nicolas

>>
>> I also have CurrentReadOnlySourceFiles undefined, but maybe I missed
>> some preamble/postscript ?
>
> There's no preamble/postscript related to CurrentReadOnlySourceFiles, it's
> in Files-ul.104. There's also an mcm to ensure the proper load order, so I
> guess it's only a problem with your image.
>
>
> Levente
>
>>
>> Nicolas
>>
>>
>
>

Levente Uzonyi-2

Re: More problems with Monticello 447

On Mon, 28 Mar 2011, Nicolas Cellier wrote:

> 2011/3/26 Levente Uzonyi <[hidden email]>:
>> On Sat, 26 Mar 2011, Nicolas Cellier wrote:
>>
>>> One of my images crashed on an infinite loop.
>>> Don't know the origin of the problem, but I had to change
>>>
>>> MCVersionName>>hash
>>> ^ (String withAll: self versionName) hash
>>
>> #versionName always returns a ByteString, so it won't be recursive. Because
>> of this, it's not worth putting MCVersionName to hashed collections, because
>> the performance will be poor.
>>
>
> OK, don't know what went on, I probably messed with load order or something.
> There was a removed (String>>versionName ^self) that might explain...
>
> However, I still consider this construction as fragile, since it
> relies on species.
> A guiding comment in MCVersionName>>species would be welcome.
> It would be even nicer to assert the implementation with a TestCase like:
>
> self assert: MCVersionName someInstance species ~= MCVersionName.
>
> Since #first: might be changed in the future and not use #species
> anymore (who knows...) ,
> then it might be even better to document the intention with a TestCase like:
>
> self assert: MCVersionName someInstance versionName class ~= MCVersionName.

I agree.

Levente

>
> Nicolas
>
>>>
>>> I also have CurrentReadOnlySourceFiles undefined, but maybe I missed
>>> some preamble/postscript ?
>>
>> There's no preamble/postscript related to CurrentReadOnlySourceFiles, it's
>> in Files-ul.104. There's also an mcm to ensure the proper load order, so I
>> guess it's only a problem with your image.
>>
>>
>> Levente
>>
>>>
>>> Nicolas
>>>
>>>
>>
>>
>
>