[Cuis] 32Mb Sources limit

> Juan,
>
> Do you have any thoughts you would like to share about on what you want do
> about this issue?
>
> I wonder if there is any documentation on the 3.11-trunk implementation of
> method trailers? Does it meet the simplicity criteria for Cuis?
>
> I have working a StandardSourceFileArrayPlus which multiplies the existing
> sourceFile pointer by 16 giving a theoretical sources/changes limit of
> 512Mb. This seems to work ok as an experiment.
>

> On 14 February 2010 16:13, keith <[hidden email]> wrote:
>> Juan,
>>
>> Do you have any thoughts you would like to share about on what you  
>> want do
>> about this issue?
>>
>> I wonder if there is any documentation on the 3.11-trunk  
>> implementation of
>> method trailers? Does it meet the simplicity criteria for Cuis?
>>
>> I have working a StandardSourceFileArrayPlus which multiplies the  
>> existing
>> sourceFile pointer by 16 giving a theoretical sources/changes limit  
>> of
>> 512Mb. This seems to work ok as an experiment.
>>
>
> New trailers imposing no limit to a file size. It can be of any size.
>
>> Keith

>
> On 14 Feb 2010, at 21:22, Igor Stasenko wrote:
>
>> On 14 February 2010 16:13, keith <[hidden email]> wrote:
>>> Juan,
>>>
>>> Do you have any thoughts you would like to share about on what you
>>> want do
>>> about this issue?
>>>
>>> I wonder if there is any documentation on the 3.11-trunk
>>> implementation of
>>> method trailers? Does it meet the simplicity criteria for Cuis?
>>>
>>> I have working a StandardSourceFileArrayPlus which multiplies the
>>> existing
>>> sourceFile pointer by 16 giving a theoretical sources/changes limit of
>>> 512Mb. This seems to work ok as an experiment.
>>>
>>
>> New trailers imposing no limit to a file size. It can be of any size.
>>
>>> Keith
>
> Thanks Igor,
>
> I understood that bit.
>
> My question is whether Juan has an opinion on your implementation with
> regards simplicity/elegance/economy and his goals for Cuis.

> keith wrote:
>>
>> On 14 Feb 2010, at 21:22, Igor Stasenko wrote:
>>
>>> On 14 February 2010 16:13, keith <[hidden email]> wrote:
>>>> Juan,
>>>>
>>>> Do you have any thoughts you would like to share about on what you
>>>> want do
>>>> about this issue?
>>>>
>>>> I wonder if there is any documentation on the 3.11-trunk
>>>> implementation of
>>>> method trailers? Does it meet the simplicity criteria for Cuis?
>>>>
>>>> I have working a StandardSourceFileArrayPlus which multiplies the
>>>> existing
>>>> sourceFile pointer by 16 giving a theoretical sources/changes limit of
>>>> 512Mb. This seems to work ok as an experiment.
>>>>
>>>
>>> New trailers imposing no limit to a file size. It can be of any size.
>>>
>>>> Keith
>>
>> Thanks Igor,
>>
>> I understood that bit.
>>
>> My question is whether Juan has an opinion on your implementation
>> with regards simplicity/elegance/economy and his goals for Cuis.
>
> Not yet. I need to actually look at it.
>
>> My use of Cuis is always building from the base start image so I am
>> unlikely to hit the limit any time soon, however the limit can be
>> extended to 512MB with about 4 methods, once the System-Sources
>> refactoring I am working on is complete (i.e. moving references to
>> RemoteString into the System-Sources package).
>
> I'd like to see that refactoring, and to load it into Cuis. Please
> tell when it is complete.
>
>> It may be simpler to use my solution, however it might be useful for
>> you to give us some indication of what the trade-offs are. Memory
>> overhead etc.
>>
>> The cost of my solution is that the sources/changes files bloat by
>> roughly 0.03% as records are padded to be saved on 16 byte boundaries.
>
> That's zero cost to me. Besides 512Mb sounds like enough not to worry
> about it anymore. On the other hand, loading Method Trailers could
> have some other uses, and it would mean syncing CompiledMethod with
> trunk, which is a good thing.
>
>> regards
>>
>> Keith
>
> Cheers,
> Juan Vuletich