Hi, Am Montag, den 22.03.2010, 21:06 +0100 schrieb Stephan Renatus: > On Sun, Mar 21, 2010 at 11:27:19PM +0100, Joachim Breitner wrote: > > > > What is the correct way to initialize the seed? > > > I was wrong about that, xscreensaver must somehow already take care of this. > (I tried adding "srandom(time(NULL))" to unicode_init() but there's an error > indicating that this shouldn't be done inside the screensaver code for > xscreensaver screensavers.) > > Now I'm left completely puzzled as I do believe that there's something weird > about the screensaver's randomness. I spent quite some time watching it and > some chars occoured way too often -- while using arc4random from above most > chars I saw where unknown to be beforehand... > > I tried a "modulo bias" workaround that was suggested here [1] using code > from here [2] and tried to capture some data with minimal examples of both > versions with the numbers from unicode-screensavers (max=3092 and 5121) and > compared their histograms but I couldn't find anything significant there, > what increases my confusion. > > [1] http://stackoverflow.com/questions/648739/objective-c-modulo-bias > [2] > http://stackoverflow.com/questions/1322510/bit-twiddling-find-next-power-of-two > > Sorry for the noise in case there actually is no "objective" bug... > Greetings, > Stephan > > PS: The workaround:
Thanks for your investigations! Just clarifying: You could not give hard evidence what, if anything, is wrong with the current code, but the patch you included did fix the impression of non-randomness you had? If that’s the case I can just apply this patch (which does not introduce a dependency) and hope that we are done with this. (Although I find it hard to believe that this can be the cause, given that ya_random() is supposed to return something between 1 and 4294967295...) Thanks, Joachim -- Joachim "nomeata" Breitner Debian Developer nome...@debian.org | ICQ# 74513189 | GPG-Keyid: 4743206C JID: nome...@joachim-breitner.de | http://people.debian.org/~nomeata
signature.asc
Description: This is a digitally signed message part
