"Andrew C. Greenberg" wrote:
Even after installing the [FIX][ENH] in Stephan's routines and recompiling the plugin, the following computations behave erratically:
16r1FFFFFFFF gcd: 16r100000000000000 16r1000000000000000/16r1000000000000000
these freeze, and the following:
16r1000000000000000//16r1000000000000000 16r1000000000000000\\16r1000000000000000
return varied results, the first tends to be either 254 or zero, the second tends to be odd negative numbers.
I cannot confirm this; for me these expressions are resulting in 16r1FFFFFFFF gcd: 16r100000000000000 1 16r1000000000000000/16r1000000000000000 1 16r1000000000000000//16r1000000000000000 1 16r1000000000000000\16r1000000000000000 0 as they should.
Did you apply changeset 'Fix_Integer.3.cs' (there have been two changesets with bug fixes after the first one...)?
A short answer: 'yes, I think so.'
I just reinstalled each of the changesets on a virgin system, built a new plugin and the results were identical to those I reported: I still can't get it to behave properly.
Note that the sample computations DO work correctly once I turn off the plugin, or with John's Long Integer stuff.
The changesets I have did not have filenames. My e-mail client did not show them as attachments, but rather as text following your messages. Accordingly, I copied them into text files and filed them in.
The "first" e-mail received from was about 75K, and entitled "[ENH] LargeIntegers plugin," containing two changesets, which I shall refer to as C1, and CInstall.
The "second" e-mail received was about 66K, and entitled "[BUG][FIX][ENH] LargeIntegers Plugin," containing a single changeset, which I shall refer to as C2.
The "third" e-mail received was about 6K, and entitled "Re: [BUG][FIX]..." containing a single changset, which I shall refer to as C3.
In filedIn C1, CInstall, C2 and C3, in order, built a plugin, installed it. SAme results as reported here.
Is it possible you made fixes not reflected in the changesets that were published? Perhaps you might send (under private mail) a "current" changeset reflecting the state of the project so we can confirm whether or not the problem is machine-specific?