08:43 <julien_> hello all. I have built openocd with gcc14 and fixed a few errors. A few inverted calloc argument order and one useless pointer check (on an array). Are you interested in a patch for those or do you prefer to have that fix by regular contributors ?

09:02 <Haohmaru> i think you should submit it, but i'm not a dev

09:29 <Dracos-Carazza> Hi, I'm trying to create a SMP configuration for a Cortex A55 octa core SoC. Is there a way to tell openocd (or the gdb?) to not set breakpoints on offline/not yet determined CPUs?

09:44 <PaulFertser> Dracos-Carazza: hi

09:44 <PaulFertser> Dracos-Carazza: basically, in the config file you add -defer-examine to the targets

09:44 <PaulFertser> And if the target is not examined no breakpoints should get set there I think?

09:45 <Dracos-Carazza> yeah, I would also expect that, but it seems to be not the case

09:46 <PaulFertser> How do you tell undesired thing happens?

09:49 <Dracos-Carazza> I get "Error: [r8a779f0.a55.1] can't add breakpoint: resource not available" and gdb doesn't seem to ignore this

09:52 <julien_> @Haohmaru thanks for your advice

09:53 <Haohmaru> julien_, mr PaulFertser should give better advice

09:54 <PaulFertser> Dracos-Carazza: but are you sure that core is not in examined state at that moment?

09:55 <PaulFertser> julien_: fixes sent to OpenOCD Gerrit are appreciated, would be nice if you send them, thank you!

09:55 <julien_> PaulFertser: thanks

09:56 <Dracos-Carazza> PaulFertser, I'm trying to debug an L4re based hypervisor with a fiasco kernel. from what I know, only core0 is active during boot until it is activated by the kernel. since I break at the entry of the bootsrapper, the kernel should not be active yet

09:56 <PaulFertser> Dracos-Carazza: yes, but I'm asking about OpenOCD point of view

09:57 <PaulFertser> Dracos-Carazza: what does "targets" say?

09:58 <Dracos-Carazza> https://beckert.dev/paste/?5848d69dee522f58#5JHDR1tidivBisP3QvR82giP7DGAhRnnq6nsdPnopFgv

10:01 <PaulFertser> Dracos-Carazza: and are you adding hardware (not software) breakpoints?

10:02 <Dracos-Carazza> yes

10:02 <PaulFertser> Dracos-Carazza: probably you might need to change src/target/breakpoints.c breakpoint_add to account for that and to skip the targets that were not examined.

10:04 <PaulFertser> Dracos-Carazza: alternatively you can use "smp off" command manually before you get everything examined, then "smp on".

10:10 <Dracos-Carazza> yeah, I'm currently trying it with a patched breakpoint.c

10:25 <Dracos-Carazza> PaulFertser, looks like it's working now, but I have to poll the cores manually in order to make them available

12:38 <julien_> PaulFertser: and the issues had already been fixed, but I'll know for next time :-)

13:25 <PaulFertser> julien_: you were building the release and not the latest commit initially?

13:25 <PaulFertser> julien_: it's not the last problem in OpenOCD so feel free to fix something else :)

13:32 <PaulFertser> Dracos-Carazza: looks like it's a shortcoming in the current implementation so if you can fix it properly and contribute that would be great.

13:55 <Dracos-Carazza> PaulFertser, I found another solution. I start debugging wihtout smp and start the debugsession. I added a breakpoint in the code after all cores have been activated via psci. When I hit the breakpoint I enable debugging for all remaining cores and switch on smp via a previously defined function

13:56 <Dracos-Carazza> this way I don't need to patch breakpoint.c

14:12 <PaulFertser> Dracos-Carazza: yes, smp can be switched on and off at any time by using that "smp" command I mentioned.

14:15 <Dracos-Carazza> yeah, I decided to do it this way, since I need to retrigger the core detection manually, after the cores have been turned on