I need some help; I'm seeing odd inconsistent behaviour when debugging userspace on a Virtex-4 (don't yet know if it's a gdbserver or kernel issue).
Here's the scenario: - tested on kernel versions: 2.6.15, 2.6.16-rc1 both w/ ml403 patches - tested with gdb versions 6.1, 6.3 & 6.4 - root file system contains: busybox-1.01, glibc-2.3.5, gdbserver - gcc version: 3.4.4 compiled with crosstool-0.38 - BDI *not* attached, and CONFIG_BDI_SWITCH is *not set* simple hello.c compiled for the target. Program loops 20 times over some printf's and simple math. On target, I start the program with: $ gdbserver 192.168.0.37:2345 ./hello Process ./hello created; pid = 268 Listening on port 2345 On the host, I connect gdb and set a breakpoint at main(), and at two places in the loop. Whenever I continue (c), I see one of the following behaviours (randomly it seems): 1. gdb stops exactly where it started from. ie. no code gets executed, 2. continue behaves properly, and execution stops at the next breakpoint, or 3. the program runs to completion, and no other breakpoints get hit. program exits normally When I do a single instruction step (si), I see either: 1. the pc does not move, or 2. the program runs freely until the next breakpoint. I've attached a gdb log with 'set debug remote 1' turned on. It looks to me that the kernel single step support is still busted on my board. It also seems to be that gdb single steps off of the previous breakpoint before reestablishing all the breakpoints and letting the process run again. Therefore, if single step is busted, then it makes sense that continue is busted to. Any help would be appreciated. Thanks, g. -- Grant Likely, B.Sc. P.Eng. Secret Lab Technologies Ltd. (403) 663-0761 -------------- next part -------------- A non-text attachment was scrubbed... Name: gda.log Type: text/x-log Size: 9506 bytes Desc: not available Url : http://ozlabs.org/pipermail/linuxppc-embedded/attachments/20060118/180a110c/attachment.bin
