search

gligli / p600fw

Teensy++ based Sequential Circuits Prophet 600 firmware remake
http://gligli.github.io/p600fw/
41 stars 23 forks source link

Sysex messages don't seem to be receiving #70

Closed snickell closed 9 years ago

snickell commented 9 years ago

I think I have the "midi tuning standard" PR updates against master, but master seems to have had a regression on receiving sysex messages. I tested against master from a year ago and sysex messages do receive correctly.

Every time I try to transmit sysex to the p600fw, I'm seeing what looks like memory corruption somewhere in midi.c (and end up getting a "sysex message overflowed").

Any idea what commits might have affected this? I'll work on tracking it down, but its the sort of bug that's a little tricky w/o jtag!

gligli commented 9 years ago

A memory corruption could occur because too much RAM is used for data and there's not enough left for stack, so do you think your changes use more RAM than master? I know we're kind of short on RAM that's why it made me think about it.

snickell commented 9 years ago

The issue occurs on unmodified gligli/master, so not something related to my changes. There's some fairly convoluted pointer arithmatic going on in midi.c, esp dealing with filling the sysex buffer, so I'm suspicious of that area of things.

Unrelated thought: I don't know the specifics of the avr architecture and how avr-gcc allocates memory, but I thought stack was usually a fixed size allocation (so from a different memory pool than malloc which means 'using more ram' shouldn't affect the stack unless you overflow the stack in a particular frame).

On Wed, Apr 22, 2015 at 2:54 AM, GliGli notifications@github.com wrote:

A memory corruption could occur because too much RAM is used for data and there's not enough left for stack, so do you think your changes use more RAM than master? I know we're kind of short on RAM that's why it made me think about it.

— Reply to this email directly or view it on GitHub https://github.com/gligli/p600fw/issues/70#issuecomment-95104664.

snickell commented 9 years ago

Sysex's receive correctly on d83dd7e (Oct 2013), but do not receive on current master b311b96 (Apr 2015). If I have time I'll try to isolate the commit that caused the regression.

gligli commented 9 years ago

Another wild guess, it could be this commit, it made the sysex buffer smaller: https://github.com/gligli/p600fw/commit/7224ec8071e5dc2ac1f2e6e1ce614ce6118c001c

snickell commented 9 years ago

I think that just changed the storage buffer, but I'll check if before/after commit made a difference.

On Wed, Apr 22, 2015 at 2:19 PM, GliGli notifications@github.com wrote:

Another wild guess, it could be this commit, it made the sysex buffer smaller: 7224ec8 https://github.com/gligli/p600fw/commit/7224ec8071e5dc2ac1f2e6e1ce614ce6118c001c

— Reply to this email directly or view it on GitHub https://github.com/gligli/p600fw/issues/70#issuecomment-95339949.

snickell commented 9 years ago

Argh, my fault. I've had my midi in/out cables reversed, but the old 'echo' kept me from ever noticing :-P Sysex working fine, my fault entirely.