| *** tpb has joined #openrisc | 00:00 | |
| *** thunor has joined #openrisc | 00:38 | |
| *** mafm has quit IRC | 00:50 | |
| *** noobineer has joined #openrisc | 03:52 | |
| *** noobineer has quit IRC | 03:54 | |
| *** rohitksingh_work has joined #openrisc | 05:24 | |
| *** heroux has quit IRC | 08:54 | |
| *** heroux has joined #openrisc | 08:56 | |
| *** Finde has quit IRC | 10:04 | |
| *** thunor has quit IRC | 12:39 | |
| *** mafm has joined #openrisc | 13:05 | |
| *** rohitksingh_work has quit IRC | 13:50 | |
| *** wtfuzz has joined #openrisc | 15:16 | |
| wtfuzz | can anyone recommend the most stable versions of OpenOCD and the or1k toolchain to use with mor1kx? | 15:23 |
|---|---|---|
| *** mafm has quit IRC | 15:28 | |
| *** mafm has joined #openrisc | 15:42 | |
| ZipCPU | wtfuxx: Ask shorne | 16:50 |
| wtfuzz | shorne: any suggestions? The git master of openocd gives CRC errors. Ubuntu version of openocd (0.10.0) works much better, but I can't seem to get a simple newlib binary to run at all, single stepping just hangs gdb. I used to have this working well but lost my environment (I created a Terasic C5G fusesoc core and ported nuttx to openrisc) | 16:58 |
| *** rohitksingh has joined #openrisc | 17:25 | |
| ZipCPU | (wtfuzz: He's on a different timezone ... you may need to wait a bit) | 17:31 |
| *** rohitksingh has quit IRC | 19:20 | |
| shorne | wtfuzz: hello, I have always been using git master | 21:19 |
| shorne | but... I admit I havent run it for a while | 21:20 |
| shorne | which version of gdb, and which SoC setup are you using? | 21:20 |
| wtfuzz | interesting. I get errors about CRC match bit when using master | 21:20 |
| wtfuzz | I actually made some progress, but I still can't single step or breakpoint | 21:21 |
| shorne | Note, I haven't probably use it in a year, I have been mailing doing iverilog sims and qemu for a while to get software stuff working | 21:21 |
| wtfuzz | gdb 7.11 is working | 21:21 |
| wtfuzz | with openocd 0.10.0 from ubuntu 18.04 .deb | 21:22 |
| shorne | ok, there are various part we maintain for the gdb to or1k debug pipeline | 21:22 |
| shorne | i.e. | 21:22 |
| shorne | gdb -> openocd -> jtag_if -> adv_debug_sys -> cpu_du -> cpu | 21:23 |
| wtfuzz | the errors from openocd would indicate a mismatch with adv_debug_sys, but I'm using the latest | 21:23 |
| shorne | there were some recent adv_debug_sys change too | 21:23 |
| wtfuzz | using altera_virtual_jtag -> adv_debug_sys -> mor1kx | 21:23 |
| wtfuzz | and wishbone master from adv_debug_sys | 21:24 |
| shorne | right, ok, let me try to reproduce | 21:25 |
| wtfuzz | hmm, I'm using olofk/adv_debug_sys which hasn't had an update in 2 years | 21:25 |
| shorne | sorry, I will not have time this morning, just have 15 minutes before work | 21:25 |
| shorne | which SoC setup are you using? | 21:25 |
| shorne | a fusesoc core? | 21:25 |
| wtfuzz | I'm using a Terasic C5G (Cyclone V GX), using my own fusesoc core on github at wtfuzz/c5g | 21:26 |
| shorne | ok, Ill try with our de0_nano setup | 21:27 |
| wtfuzz | I've at least got a basic newlib ELF blinking an LED now :) | 21:27 |
| shorne | but... debug should work | 21:27 |
| wtfuzz | yes, it should | 21:27 |
| shorne | https://github.com/olofk/de0_nano | 21:28 |
| tpb | Title: GitHub - olofk/de0_nano (at github.com) | 21:28 |
| wtfuzz | I do have some de0-nano boards, so I can build a system for one as well and try that if it helps | 21:28 |
| shorne | actually, I maybe am using my version here: https://github.com/stffrdhrn/de0_nano | 21:29 |
| tpb | Title: GitHub - stffrdhrn/de0_nano: OpenRISC SOC for the De0 Nano FPGA dev board (at github.com) | 21:29 |
| wtfuzz | which adv_debug_sys do you use? | 21:30 |
| shorne | I have set ::adv_debug_sys:3.1.0 | 21:32 |
| shorne | olof has 'adv_debug_sys', meaning just take the latest I suspect | 21:32 |
| wtfuzz | if I use the mdw command in openocd -> altera_virtual_jtag -> adv_debug_sys -> wishbone master -> RAM I still get the CRC errors, so it's something in openocd or it requires a different adv_debug_sys imo | 21:32 |
| wtfuzz | I even tried removing mor1kx completely from the design, just adv_debug_sys connected as wishbone master and on chip FPGA RAM. All kinds of errors. Older openocd works. | 21:34 |
| shorne | yes, I see, as I mentioned I havent upgraded openocd in years | 21:35 |
| shorne | I will have a look, I have to get going | 21:35 |
| wtfuzz | ok, thanks for the help | 21:35 |
| shorne | Can you put an issue on mor1kx? | 21:36 |
| wtfuzz | yep, can do | 21:36 |
| shorne | https://github.com/openrisc/mor1kx/issues | 21:36 |
| tpb | Title: Issues · openrisc/mor1kx · GitHub (at github.com) | 21:36 |
| shorne | By the way, we do have a google summer of code student working on just this issue | 21:42 |
| shorne | i.e. ensuring we have continuous integration testing of the whole debug pipeline | 21:42 |
| wtfuzz | that would be awesome :) it's currently a challenge to get a basic environment going | 21:44 |
| wtfuzz | I figured out why the newer versions of the toolchain/newlib were crashing. gcc generates ext instructions, which have to be enabled on mor1kx with .FEATURE_EXT ("ENABLED"), | 23:08 |
| wtfuzz | breakpoints and single stepping are working with GNU gdb (GDB) 8.1.50.20180317-git | 23:09 |
| wtfuzz | I can now successfully load a newlib ELF, blink an LED and printf() to the UART on my Cyclone V GX | 23:10 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!