search

mist-devel / mist-binaries

Firmware and core binaries for the MIST board
202 stars 48 forks source link

firmware_190303 breaks "load config" #60

Closed theflynn49 closed 5 years ago

theflynn49 commented 5 years ago

Hello ... This firmware release breaks the "load configuration" function from minimig-AGA 1.2.3 symptom : once a load configuration id done, OSD will not come up any more replicates exactly #24 reverting to firmware_190101 fixes the problem. Hope this helps. Rgds

gyurco commented 5 years ago

I don't see how the recent changes could affect that...will check.

gyurco commented 5 years ago

Tried it, it works for me. Load configuration reloads the core, during the reload the OSD is not available. But that's only 4-5 seconds.

theflynn49 commented 5 years ago

Thank you for your time and your support. Due to the other ticket, that called for a regression. I reinstalled firmware_190303 to try it with other minimig cores : 1.1.0, 1.2.1 and 1.2.3; they all behave the same way : loading parameters resets the core all right, but thereafter no way to get the OSD up any more.

I guess it's a weird timing with my mist hardware then. Let me close this ticket for now.

theflynn49 commented 5 years ago

I just worked on this issue. I hooked a serial console and a JTAG debugger on the arm proc. firmware_190303 breaks right in the middle of writing the console message "CPU clock : ..." during the load configuration command after the core is resetted. It doesn't make any sense unless it misbehaves during an interrupt or something.

Anyway I downloaded and recompiled the firmware, and my local version works fine. So I don't know what to do with this, it may help others to know that the image firmware_190303.upg has issues for me.

gyurco commented 5 years ago

Great you can worked it around. Actually sometimes I see different problems with different compilers with the firmware (e.g. if I compile with one version of GCC+newlib, the sort order of the files will broken), so can I ask what compiler+newlib did you use? Another problem can be the minimig core itself, there was some updates in the git repo which fixed the timing issues, maybe you can try that, too.

theflynn49 commented 5 years ago

(sorry for the delay - internet connection problems :( ) arm-none-eabi-gcc -v gcc version 6.3.1 20170620 (15:6.3.1+svn253039-1build1) dpkg -l | grep newlib ii libnewlib-arm-none-eabi 2.4.0.20160527-3 ii libstdc++-arm-none-eabi-newlib 15:6.3.1+svn253039-1+10 cat /usr/include/newlib/_newlib_version.h

define _NEWLIB_VERSION "2.4.0"

harbaum commented 5 years ago

Freel free to release your better working version. There's one thing you should test beforhand: Update to the new version and then test it's self-update capability by downgrading again. If that routine is broken then people may have a hard time updating later.

theflynn49 commented 5 years ago

Ok,