[squeak-dev] DoubleWordArray broken in Pharo
robert.withers at pm.me
Sat Mar 28 16:24:26 UTC 2020
Thanks, I'll try it out.
On 3/28/20 12:12 PM, Levente Uzonyi wrote:
> Hi Robert
> On Sat, 28 Mar 2020, Robert wrote:
>> On 3/28/20 10:16 AM, Levente Uzonyi wrote:
>>> Hi Robert,
>>> No, it's not a VM bug. My analysis I sent to you earlier is correct. It's
>>> that in Pharo DoubleWordArray is a variableWordSubclass instead of a
>>> variableDoubleWordSubclass. The latter is not available in Pharo.
>>> Here's the link to the issue they created based on my analysis you sent
>>> to them: https://github.com/pharo-project/pharo/issues/5956
>> Alright, I shared this analysis with Pharo's Discord channel. I must
>> have missed the discussion of the #variableDoubleWordSubclass:.
>>> In theory, yes, it would be possible to rewrite SHA-512 and its related
>>> hash functions to avoid using DoubleWordArray, but I see no value in doing
>> The value would be Crypto operational in Pharo. And all that follows
> Fixing DoubleWordArray in Pharo alone won't make the packages operational.
> Pharo is a moving target and it keeps breaking stuff all the time.
>> from that (ParrotTalk, SSL, SSH, Signal, Raven). Is it possible for you
>> to make these changes, please? I see this as more likely than Pharo
>> accommodating #variableDoubleWordSubclass:. And other Smalltalks...
> I've attached a changeset. If you load it before the Cryptography
> packages, CryptographyHashing will work in Pharo 8 and in the current
> Pharo 9 alpha. Some tests will fail, because random strings are generated
> with the Random class, which is different in Squeak and Pharo but still,
> everything will work.
>> k, r
>>> Pharo simply failed to properly integrate DoubleWordArray. It's there but
>>> it's disfunctional. Just wait until they fix it.
>>> P.S.: This is pretty much off topic on squeak-dev.
>>> On Sat, 28 Mar 2020, Robert wrote:
>>>> Hey Levente,
>>>> I dug a little deeper into the Pharo issue with Cryptography and it IS a VM issue, I think. Primitive 61 is failing in DoubleWordArray>>#at:put:. Pharo seems totally uninterested in fixing this so I wanted to ask you, is there any way Cryptography can avoid reliance on DoubleWordArray?
More information about the Squeak-dev