magfreq_analysis~ needs sane defaults

[jonwwilkes]

In magfreq_analysis_new you're setting x->overlap and x->winfac to zero by default. That causes a loop in the perfroutine to start from negative indices and screw up memory reads/writes when dereferencing/assigning to x->input.

In Purr Data I went ahead and checked for zero for those members and in those cases set the corresponding member to "1" to keep from crashing. If there are better default values or other fixes to this class please let me know on the issue tracker.

[ericlyon]

Nice catch. Maybe you're pulling from an earlier version. The dist I've been working from for the last few years sets default winfac to 4 and default overlap to 2 for magfreq_analysis~ in the instantiation function magfreq_analysis_new(). Those seem to be okay values for that object. For most of my FFTease objects, the defaults are overlap 8 and winfac 1. https://github.com/ericlyon/lyonpotpourri3.0/blob/master/magfreq_analysis~.c Thanks, Eric

…

On Wed, Sep 25, 2019 at 4:26 PM jonwwilkes ***@***.***> wrote: In magfreq_analysis_new you're setting x->overlap and x->winfac to zero by default. That causes a loop in the perfroutine to start from negative indices and screw up memory reads/writes when dereferencing/assigning to x->input. In Purr Data I went ahead and checked for zero for those members and in those cases set the corresponding member to "1" instead just to keep from crashing. If there are better default values or other fixes to this class please let me know on the issue tracker <https://git.purrdata.net/jwilkes/purr-data/issues>. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1?email_source=notifications&email_token=AATHYOIJNZONTVGPEF4NPLDQLPCPLA5CNFSM4I2RVOCKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HNWSNBA>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATHYOJFWSKTPFYDBOM2VKDQLPCPLANCNFSM4I2RVOCA> .

[ericlyon]

Actually got that reversed - default winfac was set to 2 and default overlap set to 4, just for magfreq_analysis~.

…

On Wed, Sep 25, 2019 at 8:56 PM Eric Lyon ***@***.***> wrote: Nice catch. Maybe you're pulling from an earlier version. The dist I've been working from for the last few years sets default winfac to 4 and default overlap to 2 for magfreq_analysis~ in the instantiation function magfreq_analysis_new(). Those seem to be okay values for that object. For most of my FFTease objects, the defaults are overlap 8 and winfac 1. https://github.com/ericlyon/lyonpotpourri3.0/blob/master/magfreq_analysis~.c Thanks, Eric On Wed, Sep 25, 2019 at 4:26 PM jonwwilkes ***@***.***> wrote: > In magfreq_analysis_new you're setting x->overlap and x->winfac to zero > by default. That causes a loop in the perfroutine to start from negative > indices and screw up memory reads/writes when dereferencing/assigning to > x->input. > > In Purr Data I went ahead and checked for zero for those members and in > those cases set the corresponding member to "1" instead just to keep from > crashing. If there are better default values or other fixes to this class > please let me know on the issue tracker > <https://git.purrdata.net/jwilkes/purr-data/issues>. > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#1?email_source=notifications&email_token=AATHYOIJNZONTVGPEF4NPLDQLPCPLA5CNFSM4I2RVOCKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HNWSNBA>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AATHYOJFWSKTPFYDBOM2VKDQLPCPLANCNFSM4I2RVOCA> > . >